vendor: OpenCV 5.0.0 snapshot at 755e50675d97db9b7d449d8bd6b09888646f6c6e

modules/xobjdetect/CMakeLists.txt

@@ -0,0 +1,6 @@
+set(the_description "Object detection algorithms")
+ocv_define_module(xobjdetect opencv_core opencv_imgproc opencv_imgcodecs opencv_features WRAP python java objc js)
+if (BUILD_opencv_apps AND NOT APPLE_FRAMEWORK)
+    add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/tools ${CMAKE_CURRENT_BINARY_DIR}/tools)
+endif()
+add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/data ${CMAKE_CURRENT_BINARY_DIR}/data)

modules/xobjdetect/README.md

@@ -0,0 +1,4 @@
+Object Detection using Boosted Features
+=======================================
+
+Uses a Waldboost cascade and local binary patterns computed as integral features for 2D object detection.

modules/xobjdetect/data/CMakeLists.txt

@@ -0,0 +1,9 @@
+file(GLOB HAAR_CASCADES haarcascades/*.xml)
+file(GLOB LBP_CASCADES lbpcascades/*.xml)
+
+install(FILES ${HAAR_CASCADES} DESTINATION ${OPENCV_OTHER_INSTALL_PATH}/haarcascades COMPONENT libs)
+install(FILES ${LBP_CASCADES}  DESTINATION ${OPENCV_OTHER_INSTALL_PATH}/lbpcascades  COMPONENT libs)
+
+if(INSTALL_TESTS AND OPENCV_TEST_DATA_PATH)
+  install(DIRECTORY "${OPENCV_TEST_DATA_PATH}/" DESTINATION "${OPENCV_TEST_DATA_INSTALL_PATH}" COMPONENT "tests")
+endif()

modules/xobjdetect/data/readme.txt

@@ -0,0 +1,7 @@
+This folder contains various data that is used by cv libraries and/or demo applications.
+----------------------------------------------------------------------------------------
+
+haarcascades - the folder contains trained classifiers for detecting objects
+               of a particular type, e.g. faces (frontal, profile), pedestrians etc.
+               Some of the classifiers have a special license - please,
+               look into the files for details.

modules/xobjdetect/doc/xobjdetect.bib

@@ -0,0 +1,56 @@
+@inproceedings{Dollar09,
+  title={Integral Channel Features.},
+  author={Doll{\'a}r, Piotr and Tu, Zhuowen and Perona, Pietro and Belongie, Serge},
+  booktitle={BMVC},
+  volume={2},
+  number={3},
+  pages={5},
+  year={2009}
+}
+
+@inproceedings{Sochman05,
+  title={Waldboost-learning for time constrained sequential detection},
+  author={Sochman, Jan and Matas, Jiri},
+  booktitle={Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on},
+  volume={2},
+  pages={150--156},
+  year={2005},
+  organization={IEEE}
+}
+
+@incollection{Liao2007,
+  title = {Learning multi-scale block local binary patterns for face recognition},
+  author = {Liao, Shengcai and Zhu, Xiangxin and Lei, Zhen and Zhang, Lun and Li, Stan Z},
+  booktitle = {Advances in Biometrics},
+  pages = {828--837},
+  year = {2007},
+  publisher = {Springer}
+}
+@article{Viola04,
+  author = {Viola, Paul and Jones, Michael J.},
+  title = {Robust real-time face detection},
+  journal = {International Journal of Computer Vision},
+  year = {2004},
+  volume = {57},
+  number = {2},
+  pages = {137--154},
+  publisher = {Kluwer Academic Publishers}
+}
+@inproceedings{Viola01,
+  author = {Viola, Paul and Jones, Michael J.},
+  title = {Rapid object detection using a boosted cascade of simple features},
+  booktitle = {Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on},
+  year = {2001},
+  pages = {I--511},
+  volume = {1},
+  publisher = {IEEE}
+}
+@inproceedings{Lienhart02,
+  author = {Lienhart, Rainer and Maydt, Jochen},
+  title = {An extended set of haar-like features for rapid object detection},
+  booktitle = {Image Processing. 2002. Proceedings. 2002 International Conference on},
+  year = {2002},
+  pages = {I--900},
+  volume = {1},
+  publisher = {IEEE}
+}

modules/xobjdetect/include/opencv2/xobjdetect.hpp

@@ -0,0 +1,719 @@
+/*
+
+By downloading, copying, installing or using the software you agree to this
+license. If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are
+disclaimed. In no event shall copyright holders or contributors be liable for
+any direct, indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+
+*/
+
+#ifndef __OPENCV_XOBJDETECT_XOBJDETECT_HPP__
+#define __OPENCV_XOBJDETECT_XOBJDETECT_HPP__
+
+#include <opencv2/core.hpp>
+#include <vector>
+#include <string>
+
+/** @defgroup xobjdetect Object Detection
+
+@{
+    @defgroup xobjdetect_cascade_classifier Cascade Classifier for Object Detection
+
+    The object detector described below has been initially proposed by Paul Viola @cite Viola01 and
+    improved by Rainer Lienhart @cite Lienhart02 .
+
+    First, a classifier (namely a *cascade of boosted classifiers working with haar-like features*) is
+    trained with a few hundred sample views of a particular object (i.e., a face or a car), called
+    positive examples, that are scaled to the same size (say, 20x20), and negative examples - arbitrary
+    images of the same size.
+
+    After a classifier is trained, it can be applied to a region of interest (of the same size as used
+    during the training) in an input image. The classifier outputs a "1" if the region is likely to show
+    the object (i.e., face/car), and "0" otherwise. To search for the object in the whole image one can
+    move the search window across the image and check every location using the classifier. The
+    classifier is designed so that it can be easily "resized" in order to be able to find the objects of
+    interest at different sizes, which is more efficient than resizing the image itself. So, to find an
+    object of an unknown size in the image the scan procedure should be done several times at different
+    scales.
+
+    The word "cascade" in the classifier name means that the resultant classifier consists of several
+    simpler classifiers (*stages*) that are applied subsequently to a region of interest until at some
+    stage the candidate is rejected or all the stages are passed. The word "boosted" means that the
+    classifiers at every stage of the cascade are complex themselves and they are built out of basic
+    classifiers using one of four different boosting techniques (weighted voting). Currently Discrete
+    Adaboost, Real Adaboost, Gentle Adaboost and Logitboost are supported. The basic classifiers are
+    decision-tree classifiers with at least 2 leaves. Haar-like features are the input to the basic
+    classifiers, and are calculated as described below. The current algorithm uses the following
+    Haar-like features:
+
+    ![image](pics/haarfeatures.png)
+
+    The feature used in a particular classifier is specified by its shape (1a, 2b etc.), position within
+    the region of interest and the scale (this scale is not the same as the scale used at the detection
+    stage, though these two scales are multiplied). For example, in the case of the third line feature
+    (2c) the response is calculated as the difference between the sum of image pixels under the
+    rectangle covering the whole feature (including the two white stripes and the black stripe in the
+    middle) and the sum of the image pixels under the black stripe multiplied by 3 in order to
+    compensate for the differences in the size of areas. The sums of pixel values over a rectangular
+    regions are calculated rapidly using integral images (see below and the integral description).
+
+    Check @ref tutorial_cascade_classifier "the corresponding tutorial" for more details.
+
+    The following reference is for the detection part only. There is a separate application called
+    opencv_traincascade that can train a cascade of boosted classifiers from a set of samples.
+
+    @note In the new C++ interface it is also possible to use LBP (local binary pattern) features in
+    addition to Haar-like features. .. [Viola01] Paul Viola and Michael J. Jones. Rapid Object Detection
+    using a Boosted Cascade of Simple Features. IEEE CVPR, 2001. The paper is available online at
+    <https://github.com/SvHey/thesis/blob/master/Literature/ObjectDetection/violaJones_CVPR2001.pdf>
+
+    @defgroup xobjdetect_hog HOG (Histogram of Oriented Gradients) descriptor and object detector
+    @defgroup xobjdetect_common Common functions and classes
+    @defgroup xobjdetect_waldboost Extended object detection
+@}
+ */
+
+typedef struct CvHaarClassifierCascade CvHaarClassifierCascade;
+
+namespace cv
+{
+
+//! @addtogroup xobjdetect_common
+//! @{
+
+///////////////////////////// Object Detection ////////////////////////////
+
+/** @brief This class is used for grouping object candidates detected by Cascade Classifier, HOG etc.
+
+instance of the class is to be passed to cv::partition
+ */
+class CV_EXPORTS SimilarRects
+{
+public:
+    SimilarRects(double _eps) : eps(_eps) {}
+    inline bool operator()(const Rect& r1, const Rect& r2) const
+    {
+        double delta = eps * ((std::min)(r1.width, r2.width) + (std::min)(r1.height, r2.height)) * 0.5;
+        return std::abs(r1.x - r2.x) <= delta &&
+            std::abs(r1.y - r2.y) <= delta &&
+            std::abs(r1.x + r1.width - r2.x - r2.width) <= delta &&
+            std::abs(r1.y + r1.height - r2.y - r2.height) <= delta;
+    }
+    double eps;
+};
+
+/** @brief Groups the object candidate rectangles.
+
+@param rectList Input/output vector of rectangles. Output vector includes retained and grouped
+rectangles. (The Python list is not modified in place.)
+@param groupThreshold Minimum possible number of rectangles minus 1. The threshold is used in a
+group of rectangles to retain it.
+@param eps Relative difference between sides of the rectangles to merge them into a group.
+
+The function is a wrapper for the generic function partition . It clusters all the input rectangles
+using the rectangle equivalence criteria that combines rectangles with similar sizes and similar
+locations. The similarity is defined by eps. When eps=0 , no clustering is done at all. If
+\f$\texttt{eps}\rightarrow +\inf\f$ , all the rectangles are put in one cluster. Then, the small
+clusters containing less than or equal to groupThreshold rectangles are rejected. In each other
+cluster, the average rectangle is computed and put into the output rectangle list.
+ */
+CV_EXPORTS   void groupRectangles(std::vector<Rect>& rectList, int groupThreshold, double eps = 0.2);
+/** @overload */
+CV_EXPORTS_W void groupRectangles(CV_IN_OUT std::vector<Rect>& rectList, CV_OUT std::vector<int>& weights,
+                                  int groupThreshold, double eps = 0.2);
+/** @overload */
+CV_EXPORTS   void groupRectangles(std::vector<Rect>& rectList, int groupThreshold,
+                                  double eps, std::vector<int>* weights, std::vector<double>* levelWeights );
+/** @overload */
+CV_EXPORTS   void groupRectangles(std::vector<Rect>& rectList, std::vector<int>& rejectLevels,
+                                  std::vector<double>& levelWeights, int groupThreshold, double eps = 0.2);
+/** @overload */
+CV_EXPORTS   void groupRectangles_meanshift(std::vector<Rect>& rectList, std::vector<double>& foundWeights,
+                                            std::vector<double>& foundScales,
+                                            double detectThreshold = 0.0, Size winDetSize = Size(64, 128));
+//! @}
+
+//! @addtogroup xobjdetect_cascade_classifier
+//! @{
+
+template<> struct DefaultDeleter<CvHaarClassifierCascade>{ CV_EXPORTS void operator ()(CvHaarClassifierCascade* obj) const; };
+
+enum { CASCADE_DO_CANNY_PRUNING    = 1,
+       CASCADE_SCALE_IMAGE         = 2,
+       CASCADE_FIND_BIGGEST_OBJECT = 4,
+       CASCADE_DO_ROUGH_SEARCH     = 8
+     };
+
+class CV_EXPORTS_W BaseCascadeClassifier : public Algorithm
+{
+public:
+    virtual ~BaseCascadeClassifier();
+    virtual bool empty() const CV_OVERRIDE = 0;
+    virtual bool load( const String& filename ) = 0;
+    virtual void detectMultiScale( InputArray image,
+                           CV_OUT std::vector<Rect>& objects,
+                           double scaleFactor,
+                           int minNeighbors, int flags,
+                           Size minSize, Size maxSize ) = 0;
+
+    virtual void detectMultiScale( InputArray image,
+                           CV_OUT std::vector<Rect>& objects,
+                           CV_OUT std::vector<int>& numDetections,
+                           double scaleFactor,
+                           int minNeighbors, int flags,
+                           Size minSize, Size maxSize ) = 0;
+
+    virtual void detectMultiScale( InputArray image,
+                                   CV_OUT std::vector<Rect>& objects,
+                                   CV_OUT std::vector<int>& rejectLevels,
+                                   CV_OUT std::vector<double>& levelWeights,
+                                   double scaleFactor,
+                                   int minNeighbors, int flags,
+                                   Size minSize, Size maxSize,
+                                   bool outputRejectLevels ) = 0;
+
+    virtual bool isOldFormatCascade() const = 0;
+    virtual Size getOriginalWindowSize() const = 0;
+    virtual int getFeatureType() const = 0;
+    virtual void* getOldCascade() = 0;
+
+    class CV_EXPORTS MaskGenerator
+    {
+    public:
+        virtual ~MaskGenerator() {}
+        virtual Mat generateMask(const Mat& src)=0;
+        virtual void initializeMask(const Mat& /*src*/) { }
+    };
+    virtual void setMaskGenerator(const Ptr<MaskGenerator>& maskGenerator) = 0;
+    virtual Ptr<MaskGenerator> getMaskGenerator() = 0;
+};
+
+/** @example samples/facedetect.cpp
+This program demonstrates usage of the Cascade classifier class
+\image html Cascade_Classifier_Tutorial_Result_Haar.jpg "Sample screenshot" width=321 height=254
+*/
+/** @brief Cascade classifier class for object detection.
+ */
+class CV_EXPORTS_W CascadeClassifier
+{
+public:
+    CV_WRAP CascadeClassifier();
+    /** @brief Loads a classifier from a file.
+
+    @param filename Name of the file from which the classifier is loaded.
+     */
+    CV_WRAP CascadeClassifier(const String& filename);
+    ~CascadeClassifier();
+    /** @brief Checks whether the classifier has been loaded.
+    */
+    CV_WRAP bool empty() const;
+    /** @brief Loads a classifier from a file.
+
+    @param filename Name of the file from which the classifier is loaded. The file may contain an old
+    HAAR classifier trained by the haartraining application or a new cascade classifier trained by the
+    traincascade application.
+     */
+    CV_WRAP bool load( const String& filename );
+    /** @brief Reads a classifier from a FileStorage node.
+
+    @note The file may contain a new cascade classifier (trained by the traincascade application) only.
+     */
+    CV_WRAP bool read( const FileNode& node );
+
+    /** @brief Detects objects of different sizes in the input image. The detected objects are returned as a list
+    of rectangles.
+
+    @param image Matrix of the type CV_8U containing an image where objects are detected.
+    @param objects Vector of rectangles where each rectangle contains the detected object, the
+    rectangles may be partially outside the original image.
+    @param scaleFactor Parameter specifying how much the image size is reduced at each image scale.
+    @param minNeighbors Parameter specifying how many neighbors each candidate rectangle should have
+    to retain it.
+    @param flags Parameter with the same meaning for an old cascade as in the function
+    cvHaarDetectObjects. It is not used for a new cascade.
+    @param minSize Minimum possible object size. Objects smaller than that are ignored.
+    @param maxSize Maximum possible object size. Objects larger than that are ignored. If `maxSize == minSize` model is evaluated on single scale.
+    */
+    CV_WRAP void detectMultiScale( InputArray image,
+                          CV_OUT std::vector<Rect>& objects,
+                          double scaleFactor = 1.1,
+                          int minNeighbors = 3, int flags = 0,
+                          Size minSize = Size(),
+                          Size maxSize = Size() );
+
+    /** @overload
+    @param image Matrix of the type CV_8U containing an image where objects are detected.
+    @param objects Vector of rectangles where each rectangle contains the detected object, the
+    rectangles may be partially outside the original image.
+    @param numDetections Vector of detection numbers for the corresponding objects. An object's number
+    of detections is the number of neighboring positively classified rectangles that were joined
+    together to form the object.
+    @param scaleFactor Parameter specifying how much the image size is reduced at each image scale.
+    @param minNeighbors Parameter specifying how many neighbors each candidate rectangle should have
+    to retain it.
+    @param flags Parameter with the same meaning for an old cascade as in the function
+    cvHaarDetectObjects. It is not used for a new cascade.
+    @param minSize Minimum possible object size. Objects smaller than that are ignored.
+    @param maxSize Maximum possible object size. Objects larger than that are ignored. If `maxSize == minSize` model is evaluated on single scale.
+    */
+    CV_WRAP_AS(detectMultiScale2) void detectMultiScale( InputArray image,
+                          CV_OUT std::vector<Rect>& objects,
+                          CV_OUT std::vector<int>& numDetections,
+                          double scaleFactor=1.1,
+                          int minNeighbors=3, int flags=0,
+                          Size minSize=Size(),
+                          Size maxSize=Size() );
+
+    /** @overload
+    This function allows you to retrieve the final stage decision certainty of classification.
+    For this, one needs to set `outputRejectLevels` on true and provide the `rejectLevels` and `levelWeights` parameter.
+    For each resulting detection, `levelWeights` will then contain the certainty of classification at the final stage.
+    This value can then be used to separate strong from weaker classifications.
+
+    A code sample on how to use it efficiently can be found below:
+    @code
+    Mat img;
+    vector<double> weights;
+    vector<int> levels;
+    vector<Rect> detections;
+    CascadeClassifier model("/path/to/your/model.xml");
+    model.detectMultiScale(img, detections, levels, weights, 1.1, 3, 0, Size(), Size(), true);
+    cerr << "Detection " << detections[0] << " with weight " << weights[0] << endl;
+    @endcode
+    */
+    CV_WRAP_AS(detectMultiScale3) void detectMultiScale( InputArray image,
+                                  CV_OUT std::vector<Rect>& objects,
+                                  CV_OUT std::vector<int>& rejectLevels,
+                                  CV_OUT std::vector<double>& levelWeights,
+                                  double scaleFactor = 1.1,
+                                  int minNeighbors = 3, int flags = 0,
+                                  Size minSize = Size(),
+                                  Size maxSize = Size(),
+                                  bool outputRejectLevels = false );
+
+    CV_WRAP bool isOldFormatCascade() const;
+    CV_WRAP Size getOriginalWindowSize() const;
+    CV_WRAP int getFeatureType() const;
+    void* getOldCascade();
+
+    CV_WRAP static bool convert(const String& oldcascade, const String& newcascade);
+
+    void setMaskGenerator(const Ptr<BaseCascadeClassifier::MaskGenerator>& maskGenerator);
+    Ptr<BaseCascadeClassifier::MaskGenerator> getMaskGenerator();
+
+    Ptr<BaseCascadeClassifier> cc;
+};
+
+CV_EXPORTS Ptr<BaseCascadeClassifier::MaskGenerator> createFaceDetectionMaskGenerator();
+//! @}
+
+//! @addtogroup xobjdetect_hog
+//! @{
+//////////////// HOG (Histogram-of-Oriented-Gradients) Descriptor and Object Detector //////////////
+
+//! struct for detection region of interest (ROI)
+struct DetectionROI
+{
+   //! scale(size) of the bounding box
+   double scale;
+   //! set of requested locations to be evaluated
+   std::vector<cv::Point> locations;
+   //! vector that will contain confidence values for each location
+   std::vector<double> confidences;
+};
+
+/**@brief Implementation of HOG (Histogram of Oriented Gradients) descriptor and object detector.
+
+the HOG descriptor algorithm introduced by Navneet Dalal and Bill Triggs @cite Dalal2005 .
+
+useful links:
+
+https://hal.inria.fr/inria-00548512/document/
+
+https://en.wikipedia.org/wiki/Histogram_of_oriented_gradients
+
+https://software.intel.com/en-us/ipp-dev-reference-histogram-of-oriented-gradients-hog-descriptor
+
+http://www.learnopencv.com/histogram-of-oriented-gradients
+
+http://www.learnopencv.com/handwritten-digits-classification-an-opencv-c-python-tutorial
+
+ */
+class CV_EXPORTS_W HOGDescriptor
+{
+public:
+    enum HistogramNormType { L2Hys = 0 //!< Default histogramNormType
+         };
+    enum { DEFAULT_NLEVELS = 64 //!< Default nlevels value.
+         };
+    enum DescriptorStorageFormat { DESCR_FORMAT_COL_BY_COL, DESCR_FORMAT_ROW_BY_ROW };
+
+    /**@brief Creates the HOG descriptor and detector with default parameters.
+
+    @param _winSize sets winSize with given value.
+    @param _blockSize sets blockSize with given value.
+    @param _blockStride sets blockStride with given value.
+    @param _cellSize sets cellSize with given value.
+    @param _nbins sets nbins with given value.
+    @param _derivAperture sets derivAperture with given value.
+    @param _winSigma sets winSigma with given value.
+    @param _histogramNormType sets histogramNormType with given value.
+    @param _L2HysThreshold sets L2HysThreshold with given value.
+    @param _gammaCorrection sets gammaCorrection with given value.
+    @param _nlevels sets nlevels with given value.
+    @param _signedGradient sets signedGradient with given value.
+    */
+    CV_WRAP HOGDescriptor(Size _winSize = Size(64, 128), Size _blockSize = Size(16, 16), Size _blockStride = Size(8, 8),
+                  Size _cellSize = Size(8, 8), int _nbins = 9, int _derivAperture = 1, double _winSigma = -1,
+                  HOGDescriptor::HistogramNormType _histogramNormType = HOGDescriptor::L2Hys,
+                  double _L2HysThreshold = 0.2, bool _gammaCorrection = true,
+                  int _nlevels = HOGDescriptor::DEFAULT_NLEVELS, bool _signedGradient = false)
+    : winSize(_winSize), blockSize(_blockSize), blockStride(_blockStride), cellSize(_cellSize),
+    nbins(_nbins), derivAperture(_derivAperture), winSigma(_winSigma),
+    histogramNormType(_histogramNormType), L2HysThreshold(_L2HysThreshold),
+    gammaCorrection(_gammaCorrection), free_coef(-1.f), nlevels(_nlevels), signedGradient(_signedGradient)
+    {}
+
+    /** @overload
+
+    Creates the HOG descriptor and detector and loads HOGDescriptor parameters and coefficients for the linear SVM classifier from a file.
+    @param filename The file name containing HOGDescriptor properties and coefficients for the linear SVM classifier.
+    */
+    CV_WRAP HOGDescriptor(const String& filename)
+    {
+        load(filename);
+    }
+
+    /** @overload
+    @param d the HOGDescriptor which cloned to create a new one.
+    */
+    HOGDescriptor(const HOGDescriptor& d)
+    {
+        d.copyTo(*this);
+    }
+
+    /**@brief Default destructor.
+    */
+    virtual ~HOGDescriptor() {}
+
+    /**@brief Returns the number of coefficients required for the classification.
+    */
+    CV_WRAP size_t getDescriptorSize() const;
+
+    /** @brief Checks if detector size equal to descriptor size.
+    */
+    CV_WRAP bool checkDetectorSize() const;
+
+    /** @brief Returns winSigma value
+    */
+    CV_WRAP double getWinSigma() const;
+
+    /**@example samples/peopledetect.cpp
+    */
+    /**@brief Sets coefficients for the linear SVM classifier.
+    @param svmdetector coefficients for the linear SVM classifier.
+    */
+    CV_WRAP virtual void setSVMDetector(InputArray svmdetector);
+
+    /** @brief Reads HOGDescriptor parameters and coefficients for the linear SVM classifier from a file node.
+    @param fn File node
+    */
+    virtual bool read(FileNode& fn);
+
+    /** @brief Stores HOGDescriptor parameters and coefficients for the linear SVM classifier in a file storage.
+    @param fs File storage
+    @param objname Object name
+    */
+    virtual void write(FileStorage& fs, const String& objname) const;
+
+    /** @brief loads HOGDescriptor parameters and coefficients for the linear SVM classifier from a file
+    @param filename Name of the file to read.
+    @param objname The optional name of the node to read (if empty, the first top-level node will be used).
+    */
+    CV_WRAP virtual bool load(const String& filename, const String& objname = String());
+
+    /** @brief saves HOGDescriptor parameters and coefficients for the linear SVM classifier to a file
+    @param filename File name
+    @param objname Object name
+    */
+    CV_WRAP virtual void save(const String& filename, const String& objname = String()) const;
+
+    /** @brief clones the HOGDescriptor
+    @param c cloned HOGDescriptor
+    */
+    virtual void copyTo(HOGDescriptor& c) const;
+
+    /** @brief Computes HOG descriptors of given image.
+    @param img Matrix of the type CV_8U containing an image where HOG features will be calculated.
+    @param descriptors Matrix of the type CV_32F
+    @param winStride Window stride. It must be a multiple of block stride.
+    @param padding Padding
+    @param locations Vector of Point
+    */
+    CV_WRAP virtual void compute(InputArray img,
+                         CV_OUT std::vector<float>& descriptors,
+                         Size winStride = Size(), Size padding = Size(),
+                         const std::vector<Point>& locations = std::vector<Point>()) const;
+
+    /** @brief Performs object detection without a multi-scale window.
+    @param img Matrix of the type CV_8U or CV_8UC3 containing an image where objects are detected.
+    @param foundLocations Vector of point where each point contains left-top corner point of detected object boundaries.
+    @param weights Vector that will contain confidence values for each detected object.
+    @param hitThreshold Threshold for the distance between features and SVM classifying plane.
+    Usually it is 0 and should be specified in the detector coefficients (as the last free coefficient).
+    But if the free coefficient is omitted (which is allowed), you can specify it manually here.
+    @param winStride Window stride. It must be a multiple of block stride.
+    @param padding Padding
+    @param searchLocations Vector of Point includes set of requested locations to be evaluated.
+    */
+    CV_WRAP virtual void detect(InputArray img, CV_OUT std::vector<Point>& foundLocations,
+                        CV_OUT std::vector<double>& weights,
+                        double hitThreshold = 0, Size winStride = Size(),
+                        Size padding = Size(),
+                        const std::vector<Point>& searchLocations = std::vector<Point>()) const;
+
+    /** @brief Performs object detection without a multi-scale window.
+    @param img Matrix of the type CV_8U or CV_8UC3 containing an image where objects are detected.
+    @param foundLocations Vector of point where each point contains left-top corner point of detected object boundaries.
+    @param hitThreshold Threshold for the distance between features and SVM classifying plane.
+    Usually it is 0 and should be specified in the detector coefficients (as the last free coefficient).
+    But if the free coefficient is omitted (which is allowed), you can specify it manually here.
+    @param winStride Window stride. It must be a multiple of block stride.
+    @param padding Padding
+    @param searchLocations Vector of Point includes locations to search.
+    */
+    virtual void detect(InputArray img, CV_OUT std::vector<Point>& foundLocations,
+                        double hitThreshold = 0, Size winStride = Size(),
+                        Size padding = Size(),
+                        const std::vector<Point>& searchLocations=std::vector<Point>()) const;
+
+    /** @brief Detects objects of different sizes in the input image. The detected objects are returned as a list
+    of rectangles.
+    @param img Matrix of the type CV_8U or CV_8UC3 containing an image where objects are detected.
+    @param foundLocations Vector of rectangles where each rectangle contains the detected object.
+    @param foundWeights Vector that will contain confidence values for each detected object.
+    @param hitThreshold Threshold for the distance between features and SVM classifying plane.
+    Usually it is 0 and should be specified in the detector coefficients (as the last free coefficient).
+    But if the free coefficient is omitted (which is allowed), you can specify it manually here.
+    @param winStride Window stride. It must be a multiple of block stride.
+    @param padding Padding
+    @param scale Coefficient of the detection window increase.
+    @param groupThreshold Coefficient to regulate the similarity threshold. When detected, some objects can be covered
+    by many rectangles. 0 means not to perform grouping.
+    @param useMeanshiftGrouping indicates grouping algorithm
+    */
+    CV_WRAP virtual void detectMultiScale(InputArray img, CV_OUT std::vector<Rect>& foundLocations,
+                                  CV_OUT std::vector<double>& foundWeights, double hitThreshold = 0,
+                                  Size winStride = Size(), Size padding = Size(), double scale = 1.05,
+                                  double groupThreshold = 2.0, bool useMeanshiftGrouping = false) const;
+
+    /** @brief Detects objects of different sizes in the input image. The detected objects are returned as a list
+    of rectangles.
+    @param img Matrix of the type CV_8U or CV_8UC3 containing an image where objects are detected.
+    @param foundLocations Vector of rectangles where each rectangle contains the detected object.
+    @param hitThreshold Threshold for the distance between features and SVM classifying plane.
+    Usually it is 0 and should be specified in the detector coefficients (as the last free coefficient).
+    But if the free coefficient is omitted (which is allowed), you can specify it manually here.
+    @param winStride Window stride. It must be a multiple of block stride.
+    @param padding Padding
+    @param scale Coefficient of the detection window increase.
+    @param groupThreshold Coefficient to regulate the similarity threshold. When detected, some objects can be covered
+    by many rectangles. 0 means not to perform grouping.
+    @param useMeanshiftGrouping indicates grouping algorithm
+    */
+    virtual void detectMultiScale(InputArray img, CV_OUT std::vector<Rect>& foundLocations,
+                                  double hitThreshold = 0, Size winStride = Size(),
+                                  Size padding = Size(), double scale = 1.05,
+                                  double groupThreshold = 2.0, bool useMeanshiftGrouping = false) const;
+
+    /** @brief  Computes gradients and quantized gradient orientations.
+    @param img Matrix contains the image to be computed
+    @param grad Matrix of type CV_32FC2 contains computed gradients
+    @param angleOfs Matrix of type CV_8UC2 contains quantized gradient orientations
+    @param paddingTL Padding from top-left
+    @param paddingBR Padding from bottom-right
+    */
+    CV_WRAP virtual void computeGradient(InputArray img, InputOutputArray grad, InputOutputArray angleOfs,
+                                 Size paddingTL = Size(), Size paddingBR = Size()) const;
+
+    /** @brief Returns coefficients of the classifier trained for people detection (for 64x128 windows).
+    */
+    CV_WRAP static std::vector<float> getDefaultPeopleDetector();
+
+    /**@example samples/hog_tapi.cpp
+    */
+    /** @brief Returns coefficients of the classifier trained for people detection (for 48x96 windows).
+    */
+    CV_WRAP static std::vector<float> getDaimlerPeopleDetector();
+
+    //! Detection window size. Align to block size and block stride. Default value is Size(64,128).
+    CV_PROP Size winSize;
+
+    //! Block size in pixels. Align to cell size. Default value is Size(16,16).
+    CV_PROP Size blockSize;
+
+    //! Block stride. It must be a multiple of cell size. Default value is Size(8,8).
+    CV_PROP Size blockStride;
+
+    //! Cell size. Default value is Size(8,8).
+    CV_PROP Size cellSize;
+
+    //! Number of bins used in the calculation of histogram of gradients. Default value is 9.
+    CV_PROP int nbins;
+
+    //! not documented
+    CV_PROP int derivAperture;
+
+    //! Gaussian smoothing window parameter.
+    CV_PROP double winSigma;
+
+    //! histogramNormType
+    CV_PROP HOGDescriptor::HistogramNormType histogramNormType;
+
+    //! L2-Hys normalization method shrinkage.
+    CV_PROP double L2HysThreshold;
+
+    //! Flag to specify whether the gamma correction preprocessing is required or not.
+    CV_PROP bool gammaCorrection;
+
+    //! coefficients for the linear SVM classifier.
+    CV_PROP std::vector<float> svmDetector;
+
+    //! coefficients for the linear SVM classifier used when OpenCL is enabled
+    UMat oclSvmDetector;
+
+    //! not documented
+    float free_coef;
+
+    //! Maximum number of detection window increases. Default value is 64
+    CV_PROP int nlevels;
+
+    //! Indicates signed gradient will be used or not
+    CV_PROP bool signedGradient;
+
+    /** @brief evaluate specified ROI and return confidence value for each location
+    @param img Matrix of the type CV_8U or CV_8UC3 containing an image where objects are detected.
+    @param locations Vector of Point
+    @param foundLocations Vector of Point where each Point is detected object's top-left point.
+    @param confidences confidences
+    @param hitThreshold Threshold for the distance between features and SVM classifying plane. Usually
+    it is 0 and should be specified in the detector coefficients (as the last free coefficient). But if
+    the free coefficient is omitted (which is allowed), you can specify it manually here
+    @param winStride winStride
+    @param padding padding
+    */
+    virtual void detectROI(InputArray img, const std::vector<cv::Point> &locations,
+                                   CV_OUT std::vector<cv::Point>& foundLocations, CV_OUT std::vector<double>& confidences,
+                                   double hitThreshold = 0, cv::Size winStride = Size(),
+                                   cv::Size padding = Size()) const;
+
+    /** @brief evaluate specified ROI and return confidence value for each location in multiple scales
+    @param img Matrix of the type CV_8U or CV_8UC3 containing an image where objects are detected.
+    @param foundLocations Vector of rectangles where each rectangle contains the detected object.
+    @param locations Vector of DetectionROI
+    @param hitThreshold Threshold for the distance between features and SVM classifying plane. Usually it is 0 and should be specified
+    in the detector coefficients (as the last free coefficient). But if the free coefficient is omitted (which is allowed), you can specify it manually here.
+    @param groupThreshold Minimum possible number of rectangles minus 1. The threshold is used in a group of rectangles to retain it.
+    */
+    virtual void detectMultiScaleROI(InputArray img,
+                                     CV_OUT std::vector<cv::Rect>& foundLocations,
+                                     std::vector<DetectionROI>& locations,
+                                     double hitThreshold = 0,
+                                     int groupThreshold = 0) const;
+
+    /** @brief Groups the object candidate rectangles.
+    @param rectList  Input/output vector of rectangles. Output vector includes retained and grouped rectangles. (The Python list is not modified in place.)
+    @param weights Input/output vector of weights of rectangles. Output vector includes weights of retained and grouped rectangles. (The Python list is not modified in place.)
+    @param groupThreshold Minimum possible number of rectangles minus 1. The threshold is used in a group of rectangles to retain it.
+    @param eps Relative difference between sides of the rectangles to merge them into a group.
+    */
+    void groupRectangles(std::vector<cv::Rect>& rectList, std::vector<double>& weights, int groupThreshold, double eps) const;
+};
+//! @}
+
+namespace xobjdetect
+{
+//! @addtogroup xobjdetect_waldboost
+//! @{
+
+
+/** @brief WaldBoost detector
+*/
+class CV_EXPORTS WBDetector {
+public:
+    /** @brief Read detector from FileNode.
+    @param node FileNode for input
+    */
+    virtual void read(const FileNode &node) = 0;
+
+    /** @brief Write detector to FileStorage.
+    @param fs FileStorage for output
+    */
+    virtual void write(FileStorage &fs) const = 0;
+
+    /** @brief Train WaldBoost detector
+    @param pos_samples Path to directory with cropped positive samples
+    @param neg_imgs Path to directory with negative (background) images
+    */
+    virtual void train(
+        const std::string& pos_samples,
+        const std::string& neg_imgs) = 0;
+
+    /** @brief Detect objects on image using WaldBoost detector
+    @param img Input image for detection
+    @param bboxes Bounding boxes coordinates output vector
+    @param confidences Confidence values for bounding boxes output vector
+    */
+    virtual void detect(
+        const Mat& img,
+        std::vector<Rect> &bboxes,
+        std::vector<double> &confidences) = 0;
+
+    /** @brief Create instance of WBDetector
+    */
+    static Ptr<WBDetector> create();
+
+    virtual ~WBDetector(){}
+};
+
+
+//! @}
+
+} /* namespace xobjdetect */
+} /* namespace cv */
+
+#include "opencv2/xobjdetect/detection_based_tracker.hpp"
+
+#endif /* __OPENCV_XOBJDETECT_XOBJDETECT_HPP__ */

modules/xobjdetect/include/opencv2/xobjdetect/detection_based_tracker.hpp

@@ -0,0 +1,222 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_XOBJDETECT_DBT_HPP
+#define OPENCV_XOBJDETECT_DBT_HPP
+
+#include <opencv2/core.hpp>
+
+#include <vector>
+
+namespace cv
+{
+
+//! @addtogroup xobjdetect_cascade_classifier
+//! @{
+
+class CV_EXPORTS DetectionBasedTracker
+{
+    public:
+        struct CV_EXPORTS Parameters
+        {
+            int maxTrackLifetime;
+            int minDetectionPeriod; //the minimal time between run of the big object detector (on the whole frame) in ms (1000 mean 1 sec), default=0
+
+            Parameters();
+        };
+
+        class IDetector
+        {
+            public:
+                IDetector():
+                    minObjSize(96, 96),
+                    maxObjSize(INT_MAX, INT_MAX),
+                    minNeighbours(2),
+                    scaleFactor(1.1f)
+                {}
+
+                virtual void detect(const cv::Mat& image, std::vector<cv::Rect>& objects) = 0;
+
+                void setMinObjectSize(const cv::Size& min)
+                {
+                    minObjSize = min;
+                }
+                void setMaxObjectSize(const cv::Size& max)
+                {
+                    maxObjSize = max;
+                }
+                cv::Size getMinObjectSize() const
+                {
+                    return minObjSize;
+                }
+                cv::Size getMaxObjectSize() const
+                {
+                    return maxObjSize;
+                }
+                float getScaleFactor()
+                {
+                    return scaleFactor;
+                }
+                void setScaleFactor(float value)
+                {
+                    scaleFactor = value;
+                }
+                int getMinNeighbours()
+                {
+                    return minNeighbours;
+                }
+                void setMinNeighbours(int value)
+                {
+                    minNeighbours = value;
+                }
+                virtual ~IDetector() {}
+
+            protected:
+                cv::Size minObjSize;
+                cv::Size maxObjSize;
+                int minNeighbours;
+                float scaleFactor;
+        };
+
+        DetectionBasedTracker(cv::Ptr<IDetector> mainDetector, cv::Ptr<IDetector> trackingDetector, const Parameters& params);
+        virtual ~DetectionBasedTracker();
+
+        virtual bool run();
+        virtual void stop();
+        virtual void resetTracking();
+
+        virtual void process(const cv::Mat& imageGray);
+
+        bool setParameters(const Parameters& params);
+        const Parameters& getParameters() const;
+
+
+        typedef std::pair<cv::Rect, int> Object;
+        virtual void getObjects(std::vector<cv::Rect>& result) const;
+        virtual void getObjects(std::vector<Object>& result) const;
+
+        enum ObjectStatus
+        {
+            DETECTED_NOT_SHOWN_YET,
+            DETECTED,
+            DETECTED_TEMPORARY_LOST,
+            WRONG_OBJECT
+        };
+        struct ExtObject
+        {
+            int id;
+            cv::Rect location;
+            ObjectStatus status;
+            ExtObject(int _id, cv::Rect _location, ObjectStatus _status)
+                :id(_id), location(_location), status(_status)
+            {
+            }
+        };
+        virtual void getObjects(std::vector<ExtObject>& result) const;
+
+
+        virtual int addObject(const cv::Rect& location); //returns id of the new object
+
+    protected:
+        class SeparateDetectionWork;
+        cv::Ptr<SeparateDetectionWork> separateDetectionWork;
+        friend void* workcycleObjectDetectorFunction(void* p);
+
+        struct InnerParameters
+        {
+            int numLastPositionsToTrack;
+            int numStepsToWaitBeforeFirstShow;
+            int numStepsToTrackWithoutDetectingIfObjectHasNotBeenShown;
+            int numStepsToShowWithoutDetecting;
+
+            float coeffTrackingWindowSize;
+            float coeffObjectSizeToTrack;
+            float coeffObjectSpeedUsingInPrediction;
+
+            InnerParameters();
+        };
+        Parameters parameters;
+        InnerParameters innerParameters;
+
+        struct TrackedObject
+        {
+            typedef std::vector<cv::Rect> PositionsVector;
+
+            PositionsVector lastPositions;
+
+            int numDetectedFrames;
+            int numFramesNotDetected;
+            int id;
+
+            TrackedObject(const cv::Rect& rect):numDetectedFrames(1), numFramesNotDetected(0)
+            {
+                lastPositions.push_back(rect);
+                id=getNextId();
+            }
+
+            static int getNextId()
+            {
+                static int _id=0;
+                return _id++;
+            }
+        };
+
+        int numTrackedSteps;
+        std::vector<TrackedObject> trackedObjects;
+
+        std::vector<float> weightsPositionsSmoothing;
+        std::vector<float> weightsSizesSmoothing;
+
+        cv::Ptr<IDetector> cascadeForTracking;
+
+        void updateTrackedObjects(const std::vector<cv::Rect>& detectedObjects);
+        cv::Rect calcTrackedObjectPositionToShow(int i) const;
+        cv::Rect calcTrackedObjectPositionToShow(int i, ObjectStatus& status) const;
+        void detectInRegion(const cv::Mat& img, const cv::Rect& r, std::vector<cv::Rect>& detectedObjectsInRegions);
+};
+
+//! @}
+
+} //end of cv namespace
+
+#endif

modules/xobjdetect/misc/java/test/CascadeClassifierTest.java

@@ -0,0 +1,104 @@
+package org.opencv.test.xobjdetect;
+
+import org.opencv.core.Mat;
+import org.opencv.core.MatOfRect;
+import org.opencv.core.Size;
+import org.opencv.imgproc.Imgproc;
+import org.opencv.xobjdetect.CascadeClassifier;
+import org.opencv.xobjdetect.Xobjdetect;
+import org.opencv.test.OpenCVTestCase;
+import org.opencv.test.OpenCVTestRunner;
+
+public class CascadeClassifierTest extends OpenCVTestCase {
+
+    private CascadeClassifier cc;
+
+    @Override
+    protected void setUp() throws Exception {
+        super.setUp();
+
+        cc = null;
+    }
+
+    public void testCascadeClassifier() {
+        cc = new CascadeClassifier();
+        assertNotNull(cc);
+    }
+
+    public void testCascadeClassifierString() {
+        cc = new CascadeClassifier(OpenCVTestRunner.LBPCASCADE_FRONTALFACE_PATH);
+        assertNotNull(cc);
+    }
+
+    public void testDetectMultiScaleMatListOfRect() {
+        CascadeClassifier cc = new CascadeClassifier(OpenCVTestRunner.LBPCASCADE_FRONTALFACE_PATH);
+        MatOfRect faces = new MatOfRect();
+
+        Mat greyLena = new Mat();
+        Imgproc.cvtColor(rgbLena, greyLena, Imgproc.COLOR_RGB2GRAY);
+        Imgproc.equalizeHist(greyLena, greyLena);
+
+        cc.detectMultiScale(greyLena, faces, 1.1, 3, Xobjdetect.CASCADE_SCALE_IMAGE, new Size(30, 30), new Size());
+        assertEquals(1, faces.total());
+    }
+
+    public void testDetectMultiScaleMatListOfRectDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleIntInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleIntIntSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleIntIntSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfIntegerListOfDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfIntegerListOfDoubleDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfIntegerListOfDoubleDoubleInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfIntegerListOfDoubleDoubleIntInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfIntegerListOfDoubleDoubleIntIntSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfIntegerListOfDoubleDoubleIntIntSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfIntegerListOfDoubleDoubleIntIntSizeSizeBoolean() {
+        fail("Not yet implemented");
+    }
+
+    public void testEmpty() {
+        cc = new CascadeClassifier();
+        assertTrue(cc.empty());
+    }
+
+    public void testLoad() {
+        cc = new CascadeClassifier();
+        cc.load(OpenCVTestRunner.LBPCASCADE_FRONTALFACE_PATH);
+        assertFalse(cc.empty());
+    }
+
+}

modules/xobjdetect/misc/java/test/HOGDescriptorTest.java

@@ -0,0 +1,259 @@
+package org.opencv.test.xobjdetect;
+
+import org.opencv.xobjdetect.HOGDescriptor;
+import org.opencv.test.OpenCVTestCase;
+
+public class HOGDescriptorTest extends OpenCVTestCase {
+
+    public void testCheckDetectorSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testComputeGradientMatMatMat() {
+        fail("Not yet implemented");
+    }
+
+    public void testComputeGradientMatMatMatSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testComputeGradientMatMatMatSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testComputeMatListOfFloat() {
+        fail("Not yet implemented");
+    }
+
+    public void testComputeMatListOfFloatSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testComputeMatListOfFloatSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testComputeMatListOfFloatSizeSizeListOfPoint() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPoint() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointDoubleSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointDoubleSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointDoubleSizeSizeListOfPoint() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointListOfDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointListOfDoubleDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointListOfDoubleDoubleSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointListOfDoubleDoubleSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMatListOfPointListOfDoubleDoubleSizeSizeListOfPoint() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRect() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleSizeSizeDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleSizeSizeDoubleDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectDoubleSizeSizeDoubleDoubleBoolean() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfDoubleDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfDoubleDoubleSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfDoubleDoubleSizeSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfDoubleDoubleSizeSizeDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfDoubleDoubleSizeSizeDoubleDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testDetectMultiScaleMatListOfRectListOfDoubleDoubleSizeSizeDoubleDoubleBoolean() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_blockSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_blockStride() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_cellSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_derivAperture() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_gammaCorrection() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_histogramNormType() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_L2HysThreshold() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_nbins() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_nlevels() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_svmDetector() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_winSigma() {
+        fail("Not yet implemented");
+    }
+
+    public void testGet_winSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testGetDaimlerPeopleDetector() {
+        fail("Not yet implemented");
+    }
+
+    public void testGetDefaultPeopleDetector() {
+        fail("Not yet implemented");
+    }
+
+    public void testGetDescriptorSize() {
+        fail("Not yet implemented");
+    }
+
+    public void testGetWinSigma() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptor() {
+        HOGDescriptor hog = new HOGDescriptor();
+
+        assertNotNull(hog);
+        assertEquals(HOGDescriptor.DEFAULT_NLEVELS, hog.get_nlevels());
+    }
+
+    public void testHOGDescriptorSizeSizeSizeSizeInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptorSizeSizeSizeSizeIntInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptorSizeSizeSizeSizeIntIntDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptorSizeSizeSizeSizeIntIntDoubleInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptorSizeSizeSizeSizeIntIntDoubleIntDouble() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptorSizeSizeSizeSizeIntIntDoubleIntDoubleBoolean() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptorSizeSizeSizeSizeIntIntDoubleIntDoubleBooleanInt() {
+        fail("Not yet implemented");
+    }
+
+    public void testHOGDescriptorString() {
+        fail("Not yet implemented");
+    }
+
+    public void testLoadString() {
+        fail("Not yet implemented");
+    }
+
+    public void testLoadStringString() {
+        fail("Not yet implemented");
+    }
+
+    public void testSaveString() {
+        fail("Not yet implemented");
+    }
+
+    public void testSaveStringString() {
+        fail("Not yet implemented");
+    }
+
+    public void testSetSVMDetector() {
+        fail("Not yet implemented");
+    }
+
+}

modules/xobjdetect/misc/java/test/ObjdetectTest.java

@@ -0,0 +1,42 @@
+package org.opencv.test.xobjdetect;
+
+import org.opencv.test.OpenCVTestCase;
+
+public class ObjdetectTest extends OpenCVTestCase {
+
+    public void testGroupRectanglesListOfRectListOfIntegerInt() {
+        fail("Not yet implemented");
+        /*
+        final int NUM = 10;
+        MatOfRect rects = new MatOfRect();
+        rects.alloc(NUM);
+
+        for (int i = 0; i < NUM; i++)
+            rects.put(i, 0, 10, 10, 20, 20);
+
+        int groupThreshold = 1;
+        Objdetect.groupRectangles(rects, null, groupThreshold);//TODO: second parameter should not be null
+        assertEquals(1, rects.total());
+        */
+    }
+
+    public void testGroupRectanglesListOfRectListOfIntegerIntDouble() {
+        fail("Not yet implemented");
+        /*
+        final int NUM = 10;
+        MatOfRect rects = new MatOfRect();
+        rects.alloc(NUM);
+
+        for (int i = 0; i < NUM; i++)
+            rects.put(i, 0, 10, 10, 20, 20);
+
+        for (int i = 0; i < NUM; i++)
+            rects.put(i, 0, 10, 10, 25, 25);
+
+        int groupThreshold = 1;
+        double eps = 0.2;
+        Objdetect.groupRectangles(rects, null, groupThreshold, eps);//TODO: second parameter should not be null
+        assertEquals(2, rects.size());
+        */
+    }
+}

modules/xobjdetect/misc/python/pyopencv_xobjdetect.hpp

@@ -0,0 +1,8 @@
+#ifdef HAVE_OPENCV_XOBJDETECT
+
+#include "opencv2/xobjdetect.hpp"
+
+typedef HOGDescriptor::HistogramNormType HOGDescriptor_HistogramNormType;
+typedef HOGDescriptor::DescriptorStorageFormat HOGDescriptor_DescriptorStorageFormat;
+
+#endif

modules/xobjdetect/misc/python/test/test_facedetect.py

@@ -0,0 +1,92 @@
+#!/usr/bin/env python
+
+'''
+face detection using haar cascades
+'''
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+
+def detect(img, cascade):
+    rects = cascade.detectMultiScale(img, scaleFactor=1.275, minNeighbors=4, minSize=(30, 30),
+                                     flags=cv.CASCADE_SCALE_IMAGE)
+    if len(rects) == 0:
+        return []
+    rects[:,2:] += rects[:,:2]
+    return rects
+
+from tests_common import NewOpenCVTests, intersectionRate
+
+class facedetect_test(NewOpenCVTests):
+
+    def test_facedetect(self):
+        cascade_fn = self.extraTestDataPath + '/cv/cascadeandhog/cascades/haarcascade_frontalface_alt.xml'
+        nested_fn  = self.extraTestDataPath + '/cv/cascadeandhog/cascades/haarcascade_eye.xml'
+
+        cascade = cv.CascadeClassifier(cascade_fn)
+        nested = cv.CascadeClassifier(nested_fn)
+
+        samples = ['samples/data/lena.jpg', 'cv/cascadeandhog/images/mona-lisa.png']
+
+        faces = []
+        eyes = []
+
+        testFaces = [
+        #lena
+        [[218, 200, 389, 371],
+        [ 244, 240, 294, 290],
+        [ 309, 246, 352, 289]],
+
+        #lisa
+        [[167, 119, 307, 259],
+        [188, 153, 229, 194],
+        [236, 153, 277, 194]]
+        ]
+
+        for sample in samples:
+
+            img = self.get_sample(  sample)
+            gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
+            gray = cv.GaussianBlur(gray, (5, 5), 0)
+
+            rects = detect(gray, cascade)
+            faces.append(rects)
+
+            if not nested.empty():
+                for x1, y1, x2, y2 in rects:
+                    roi = gray[y1:y2, x1:x2]
+                    subrects = detect(roi.copy(), nested)
+
+                    for rect in subrects:
+                        rect[0] += x1
+                        rect[2] += x1
+                        rect[1] += y1
+                        rect[3] += y1
+
+                    eyes.append(subrects)
+
+        faces_matches = 0
+        eyes_matches = 0
+
+        eps = 0.8
+
+        for i in range(len(faces)):
+            for j in range(len(testFaces)):
+                if intersectionRate(faces[i][0], testFaces[j][0]) > eps:
+                    faces_matches += 1
+                    #check eyes
+                    if len(eyes[i]) == 2:
+                        if intersectionRate(eyes[i][0], testFaces[j][1]) > eps and intersectionRate(eyes[i][1] , testFaces[j][2]) > eps:
+                            eyes_matches += 1
+                        elif intersectionRate(eyes[i][1], testFaces[j][1]) > eps and intersectionRate(eyes[i][0], testFaces[j][2]) > eps:
+                            eyes_matches += 1
+
+        self.assertEqual(faces_matches, 2)
+        self.assertEqual(eyes_matches, 2)
+
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/xobjdetect/misc/python/test/test_peopledetect.py

@@ -0,0 +1,65 @@
+#!/usr/bin/env python
+
+'''
+example to detect upright people in images using HOG features
+'''
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+
+
+def inside(r, q):
+    rx, ry, rw, rh = r
+    qx, qy, qw, qh = q
+    return rx > qx and ry > qy and rx + rw < qx + qw and ry + rh < qy + qh
+
+from tests_common import NewOpenCVTests, intersectionRate
+
+class peopledetect_test(NewOpenCVTests):
+    def test_peopledetect(self):
+
+        hog = cv.HOGDescriptor( (48, 96) )
+        hog.setSVMDetector( cv.HOGDescriptor_getDaimlerPeopleDetector() )
+
+        dirPath = 'samples/data/'
+        samples = ['basketball1.png', 'basketball2.png']
+
+        testPeople = [
+        [[23, 76, 164, 477], [440, 22, 637, 478]],
+        [[23, 76, 164, 477], [440, 22, 637, 478]]
+        ]
+
+        eps = 0.5
+
+        for sample in samples:
+
+            img = self.get_sample(dirPath + sample, 0)
+
+            found, _w = hog.detectMultiScale(img, winStride=(8,8), padding=(32,32), scale=1.05)
+            found_filtered = []
+            for ri, r in enumerate(found):
+                for qi, q in enumerate(found):
+                    if ri != qi and inside(r, q):
+                        break
+                else:
+                    found_filtered.append(r)
+
+            matches = 0
+
+            for i in range(len(found_filtered)):
+                for j in range(len(testPeople)):
+
+                    found_rect = (found_filtered[i][0], found_filtered[i][1],
+                        found_filtered[i][0] + found_filtered[i][2],
+                        found_filtered[i][1] + found_filtered[i][3])
+
+                    if intersectionRate(found_rect, testPeople[j][0]) > eps or intersectionRate(found_rect, testPeople[j][1]) > eps:
+                        matches += 1
+
+            self.assertGreater(matches, 0)
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/xobjdetect/perf/opencl/perf_cascades.cpp

@@ -0,0 +1,61 @@
+#include "../perf_precomp.hpp"
+#include <opencv2/imgproc.hpp>
+
+#include "opencv2/ts/ocl_perf.hpp"
+
+namespace opencv_test
+{
+using namespace perf;
+
+typedef tuple<std::string, std::string, int> Cascade_Image_MinSize_t;
+typedef perf::TestBaseWithParam<Cascade_Image_MinSize_t> Cascade_Image_MinSize;
+
+#ifdef HAVE_OPENCL
+
+OCL_PERF_TEST_P(Cascade_Image_MinSize, CascadeClassifier,
+                 testing::Combine(
+                    testing::Values( string("cv/cascadeandhog/cascades/haarcascade_frontalface_alt.xml"),
+                                     string("cv/cascadeandhog/cascades/haarcascade_frontalface_alt2.xml"),
+                                     string("cv/cascadeandhog/cascades/lbpcascade_frontalface.xml") ),
+                    testing::Values( string("cv/shared/lena.png"),
+                                     string("cv/cascadeandhog/images/bttf301.png"),
+                                     string("cv/cascadeandhog/images/class57.png") ),
+                    testing::Values(30, 64, 90) ) )
+{
+    const string cascadePath = get<0>(GetParam());
+    const string imagePath   = get<1>(GetParam());
+    int min_size = get<2>(GetParam());
+    Size minSize(min_size, min_size);
+
+    CascadeClassifier cc( getDataPath(cascadePath) );
+    if (cc.empty())
+        FAIL() << "Can't load cascade file: " << getDataPath(cascadePath);
+
+    Mat img = imread(getDataPath(imagePath), IMREAD_GRAYSCALE);
+    if (img.empty())
+        FAIL() << "Can't load source image: " << getDataPath(imagePath);
+
+    vector<Rect> faces;
+
+    equalizeHist(img, img);
+    declare.in(img).time(60);
+
+    UMat uimg = img.getUMat(ACCESS_READ);
+
+    while(next())
+    {
+        faces.clear();
+        cvtest::ocl::perf::safeFinish();
+
+        startTimer();
+        cc.detectMultiScale(uimg, faces, 1.1, 3, 0, minSize);
+        stopTimer();
+    }
+
+    sort(faces.begin(), faces.end(), comparators::RectLess());
+    SANITY_CHECK(faces, min_size/5);
+}
+
+#endif //HAVE_OPENCL
+
+} // namespace

modules/xobjdetect/perf/opencl/perf_hogdetect.cpp

@@ -0,0 +1,93 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Fangfang Bai, fangfang@multicorewareinc.com
+//    Jin Ma,       jin@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "../perf_precomp.hpp"
+#include "opencv2/ts/ocl_perf.hpp"
+
+#ifdef HAVE_OPENCL
+
+namespace opencv_test {
+namespace ocl {
+///////////// HOG////////////////////////
+
+struct RectLess
+{
+    bool operator()(const cv::Rect& a,
+        const cv::Rect& b) const
+    {
+        if (a.x != b.x)
+            return a.x < b.x;
+        else if (a.y != b.y)
+            return a.y < b.y;
+        else if (a.width != b.width)
+            return a.width < b.width;
+        else
+            return a.height < b.height;
+    }
+};
+
+OCL_PERF_TEST(HOGFixture, HOG)
+{
+    UMat src;
+    imread(getDataPath("gpu/hog/road.png"), cv::IMREAD_GRAYSCALE).copyTo(src);
+    ASSERT_FALSE(src.empty());
+
+    vector<cv::Rect> found_locations;
+    declare.in(src);
+
+    HOGDescriptor hog;
+    hog.setSVMDetector(hog.getDefaultPeopleDetector());
+
+    OCL_TEST_CYCLE() hog.detectMultiScale(src, found_locations);
+
+    std::sort(found_locations.begin(), found_locations.end(), RectLess());
+    SANITY_CHECK(found_locations, 3);
+}
+
+}
+}
+
+#endif

modules/xobjdetect/perf/perf_main.cpp

@@ -0,0 +1,7 @@
+#include "perf_precomp.hpp"
+
+#if defined(HAVE_HPX)
+    #include <hpx/hpx_main.hpp>
+#endif
+
+CV_PERF_TEST_MAIN(xobjdetect)

modules/xobjdetect/perf/perf_precomp.hpp

@@ -0,0 +1,7 @@
+#ifndef __OPENCV_PERF_PRECOMP_HPP__
+#define __OPENCV_PERF_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/xobjdetect.hpp"
+
+#endif

modules/xobjdetect/samples/dbt_face_detection.cpp

@@ -0,0 +1,111 @@
+#if defined(__linux__) || defined(LINUX) || defined(__APPLE__) || defined(ANDROID) || (defined(_MSC_VER) && _MSC_VER>=1800)
+
+#include <opencv2/imgproc.hpp>  // Gaussian Blur
+#include <opencv2/core.hpp>        // Basic OpenCV structures (cv::Mat, Scalar)
+#include <opencv2/videoio.hpp>
+#include <opencv2/highgui.hpp>  // OpenCV window I/O
+#include <opencv2/features.hpp>
+#include <opencv2/xobjdetect.hpp>
+
+#include <stdio.h>
+
+using namespace std;
+using namespace cv;
+
+const string WindowName = "Face Detection example";
+
+class CascadeDetectorAdapter: public DetectionBasedTracker::IDetector
+{
+    public:
+        CascadeDetectorAdapter(cv::Ptr<cv::CascadeClassifier> detector):
+            IDetector(),
+            Detector(detector)
+        {
+            CV_Assert(detector);
+        }
+
+        void detect(const cv::Mat &Image, std::vector<cv::Rect> &objects) CV_OVERRIDE
+        {
+            Detector->detectMultiScale(Image, objects, scaleFactor, minNeighbours, 0, minObjSize, maxObjSize);
+        }
+
+        virtual ~CascadeDetectorAdapter() CV_OVERRIDE
+        {}
+
+    private:
+        CascadeDetectorAdapter();
+        cv::Ptr<cv::CascadeClassifier> Detector;
+ };
+
+int main(int , char** )
+{
+    namedWindow(WindowName);
+
+    VideoCapture VideoStream(0);
+
+    if (!VideoStream.isOpened())
+    {
+        printf("Error: Cannot open video stream from camera\n");
+        return 1;
+    }
+
+    std::string cascadeFrontalfilename = samples::findFile("data/lbpcascades/lbpcascade_frontalface.xml");
+    cv::Ptr<cv::CascadeClassifier> cascade = makePtr<cv::CascadeClassifier>(cascadeFrontalfilename);
+    cv::Ptr<DetectionBasedTracker::IDetector> MainDetector = makePtr<CascadeDetectorAdapter>(cascade);
+    if ( cascade->empty() )
+    {
+      printf("Error: Cannot load %s\n", cascadeFrontalfilename.c_str());
+      return 2;
+    }
+
+    cascade = makePtr<cv::CascadeClassifier>(cascadeFrontalfilename);
+    cv::Ptr<DetectionBasedTracker::IDetector> TrackingDetector = makePtr<CascadeDetectorAdapter>(cascade);
+    if ( cascade->empty() )
+    {
+      printf("Error: Cannot load %s\n", cascadeFrontalfilename.c_str());
+      return 2;
+    }
+
+    DetectionBasedTracker::Parameters params;
+    DetectionBasedTracker Detector(MainDetector, TrackingDetector, params);
+
+    if (!Detector.run())
+    {
+        printf("Error: Detector initialization failed\n");
+        return 2;
+    }
+
+    Mat ReferenceFrame;
+    Mat GrayFrame;
+    vector<Rect> Faces;
+
+    do
+    {
+        VideoStream >> ReferenceFrame;
+        cvtColor(ReferenceFrame, GrayFrame, COLOR_BGR2GRAY);
+        Detector.process(GrayFrame);
+        Detector.getObjects(Faces);
+
+        for (size_t i = 0; i < Faces.size(); i++)
+        {
+            rectangle(ReferenceFrame, Faces[i], Scalar(0,255,0));
+        }
+
+        imshow(WindowName, ReferenceFrame);
+    } while (waitKey(30) < 0);
+
+    Detector.stop();
+
+    return 0;
+}
+
+#else
+
+#include <stdio.h>
+int main()
+{
+    printf("This sample works for UNIX or ANDROID or Visual Studio 2013+ only\n");
+    return 0;
+}
+
+#endif

modules/xobjdetect/samples/facedetect.cpp

@@ -0,0 +1,257 @@
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/videoio.hpp"
+#include <iostream>
+
+using namespace std;
+using namespace cv;
+
+static void help(const char** argv)
+{
+    cout << "\nThis program demonstrates the use of cv::CascadeClassifier class to detect objects (Face + eyes). You can use Haar or LBP features.\n"
+            "This classifier can recognize many kinds of rigid objects, once the appropriate classifier is trained.\n"
+            "It's most known use is for faces.\n"
+            "Usage:\n"
+        <<  argv[0]
+        <<  "   [--cascade=<cascade_path> this is the primary trained classifier such as frontal face]\n"
+            "   [--nested-cascade[=nested_cascade_path this an optional secondary classifier such as eyes]]\n"
+            "   [--scale=<image scale greater or equal to 1, try 1.3 for example>]\n"
+            "   [--try-flip]\n"
+            "   [filename|camera_index]\n\n"
+            "example:\n"
+        <<  argv[0]
+        <<  " --cascade=\"data/haarcascades/haarcascade_frontalface_alt.xml\" --nested-cascade=\"data/haarcascades/haarcascade_eye_tree_eyeglasses.xml\" --scale=1.3\n\n"
+            "During execution:\n\tHit any key to quit.\n"
+            "\tUsing OpenCV version " << CV_VERSION << "\n" << endl;
+}
+
+void detectAndDraw( Mat& img, CascadeClassifier& cascade,
+                    CascadeClassifier& nestedCascade,
+                    double scale, bool tryflip );
+
+string cascadeName;
+string nestedCascadeName;
+
+int main( int argc, const char** argv )
+{
+    VideoCapture capture;
+    Mat frame, image;
+    string inputName;
+    bool tryflip;
+    CascadeClassifier cascade, nestedCascade;
+    double scale;
+
+    cv::CommandLineParser parser(argc, argv,
+        "{help h||}"
+        "{cascade|data/haarcascades/haarcascade_frontalface_alt.xml|}"
+        "{nested-cascade|data/haarcascades/haarcascade_eye_tree_eyeglasses.xml|}"
+        "{scale|1|}{try-flip||}{@filename||}"
+    );
+    if (parser.has("help"))
+    {
+        help(argv);
+        return 0;
+    }
+    cascadeName = parser.get<string>("cascade");
+    nestedCascadeName = parser.get<string>("nested-cascade");
+    scale = parser.get<double>("scale");
+    if (scale < 1)
+        scale = 1;
+    tryflip = parser.has("try-flip");
+    inputName = parser.get<string>("@filename");
+    if (!parser.check())
+    {
+        parser.printErrors();
+        return 0;
+    }
+    if (!nestedCascade.load(samples::findFileOrKeep(nestedCascadeName)))
+        cerr << "WARNING: Could not load classifier cascade for nested objects" << endl;
+    if (!cascade.load(samples::findFile(cascadeName)))
+    {
+        cerr << "ERROR: Could not load classifier cascade" << endl;
+        help(argv);
+        return -1;
+    }
+    if( inputName.empty() || (isdigit(inputName[0]) && inputName.size() == 1) )
+    {
+        int camera = inputName.empty() ? 0 : inputName[0] - '0';
+        if(!capture.open(camera))
+        {
+            cout << "Capture from camera #" <<  camera << " didn't work" << endl;
+            return 1;
+        }
+    }
+    else if (!inputName.empty())
+    {
+        image = imread(samples::findFileOrKeep(inputName), IMREAD_COLOR);
+        if (image.empty())
+        {
+            if (!capture.open(samples::findFileOrKeep(inputName)))
+            {
+                cout << "Could not read " << inputName << endl;
+                return 1;
+            }
+        }
+    }
+    else
+    {
+        image = imread(samples::findFile("lena.jpg"), IMREAD_COLOR);
+        if (image.empty())
+        {
+            cout << "Couldn't read lena.jpg" << endl;
+            return 1;
+        }
+    }
+
+    if( capture.isOpened() )
+    {
+        cout << "Video capturing has been started ..." << endl;
+
+        for(;;)
+        {
+            capture >> frame;
+            if( frame.empty() )
+                break;
+
+            Mat frame1 = frame.clone();
+            detectAndDraw( frame1, cascade, nestedCascade, scale, tryflip );
+
+            char c = (char)waitKey(10);
+            if( c == 27 || c == 'q' || c == 'Q' )
+                break;
+        }
+    }
+    else
+    {
+        cout << "Detecting face(s) in " << inputName << endl;
+        if( !image.empty() )
+        {
+            detectAndDraw( image, cascade, nestedCascade, scale, tryflip );
+            waitKey(0);
+        }
+        else if( !inputName.empty() )
+        {
+            /* assume it is a text file containing the
+            list of the image filenames to be processed - one per line */
+            FILE* f = fopen( inputName.c_str(), "rt" );
+            if( f )
+            {
+                char buf[1000+1];
+                while( fgets( buf, 1000, f ) )
+                {
+                    int len = (int)strlen(buf);
+                    while( len > 0 && isspace(buf[len-1]) )
+                        len--;
+                    buf[len] = '\0';
+                    cout << "file " << buf << endl;
+                    image = imread( buf, IMREAD_COLOR );
+                    if( !image.empty() )
+                    {
+                        detectAndDraw( image, cascade, nestedCascade, scale, tryflip );
+                        char c = (char)waitKey(0);
+                        if( c == 27 || c == 'q' || c == 'Q' )
+                            break;
+                    }
+                    else
+                    {
+                        cerr << "Aw snap, couldn't read image " << buf << endl;
+                    }
+                }
+                fclose(f);
+            }
+        }
+    }
+
+    return 0;
+}
+
+void detectAndDraw( Mat& img, CascadeClassifier& cascade,
+                    CascadeClassifier& nestedCascade,
+                    double scale, bool tryflip )
+{
+    double t = 0;
+    vector<Rect> faces, faces2;
+    const static Scalar colors[] =
+    {
+        Scalar(255,0,0),
+        Scalar(255,128,0),
+        Scalar(255,255,0),
+        Scalar(0,255,0),
+        Scalar(0,128,255),
+        Scalar(0,255,255),
+        Scalar(0,0,255),
+        Scalar(255,0,255)
+    };
+    Mat gray, smallImg;
+
+    cvtColor( img, gray, COLOR_BGR2GRAY );
+    double fx = 1 / scale;
+    resize( gray, smallImg, Size(), fx, fx, INTER_LINEAR_EXACT );
+    equalizeHist( smallImg, smallImg );
+
+    t = (double)getTickCount();
+    cascade.detectMultiScale( smallImg, faces,
+        1.1, 2, 0
+        //|CASCADE_FIND_BIGGEST_OBJECT
+        //|CASCADE_DO_ROUGH_SEARCH
+        |CASCADE_SCALE_IMAGE,
+        Size(30, 30) );
+    if( tryflip )
+    {
+        flip(smallImg, smallImg, 1);
+        cascade.detectMultiScale( smallImg, faces2,
+                                 1.1, 2, 0
+                                 //|CASCADE_FIND_BIGGEST_OBJECT
+                                 //|CASCADE_DO_ROUGH_SEARCH
+                                 |CASCADE_SCALE_IMAGE,
+                                 Size(30, 30) );
+        for( vector<Rect>::const_iterator r = faces2.begin(); r != faces2.end(); ++r )
+        {
+            faces.push_back(Rect(smallImg.cols - r->x - r->width, r->y, r->width, r->height));
+        }
+    }
+    t = (double)getTickCount() - t;
+    printf( "detection time = %g ms\n", t*1000/getTickFrequency());
+    for ( size_t i = 0; i < faces.size(); i++ )
+    {
+        Rect r = faces[i];
+        Mat smallImgROI;
+        vector<Rect> nestedObjects;
+        Point center;
+        Scalar color = colors[i%8];
+        int radius;
+
+        double aspect_ratio = (double)r.width/r.height;
+        if( 0.75 < aspect_ratio && aspect_ratio < 1.3 )
+        {
+            center.x = cvRound((r.x + r.width*0.5)*scale);
+            center.y = cvRound((r.y + r.height*0.5)*scale);
+            radius = cvRound((r.width + r.height)*0.25*scale);
+            circle( img, center, radius, color, 3, 8, 0 );
+        }
+        else
+            rectangle( img, Point(cvRound(r.x*scale), cvRound(r.y*scale)),
+                       Point(cvRound((r.x + r.width-1)*scale), cvRound((r.y + r.height-1)*scale)),
+                       color, 3, 8, 0);
+        if( nestedCascade.empty() )
+            continue;
+        smallImgROI = smallImg( r );
+        nestedCascade.detectMultiScale( smallImgROI, nestedObjects,
+            1.1, 2, 0
+            //|CASCADE_FIND_BIGGEST_OBJECT
+            //|CASCADE_DO_ROUGH_SEARCH
+            //|CASCADE_DO_CANNY_PRUNING
+            |CASCADE_SCALE_IMAGE,
+            Size(30, 30) );
+        for ( size_t j = 0; j < nestedObjects.size(); j++ )
+        {
+            Rect nr = nestedObjects[j];
+            center.x = cvRound((r.x + nr.x + nr.width*0.5)*scale);
+            center.y = cvRound((r.y + nr.y + nr.height*0.5)*scale);
+            radius = cvRound((nr.width + nr.height)*0.25*scale);
+            circle( img, center, radius, color, 3, 8, 0 );
+        }
+    }
+    imshow( "result", img );
+}

modules/xobjdetect/samples/facial_features.cpp

@@ -0,0 +1,215 @@
+/*
+ * Author: Samyak Datta (datta[dot]samyak[at]gmail.com)
+ *
+ * A program to detect facial feature points using
+ * Haarcascade classifiers for face, eyes, nose and mouth
+ *
+ */
+
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+
+#include <iostream>
+#include <cstdio>
+#include <vector>
+#include <algorithm>
+
+using namespace std;
+using namespace cv;
+
+// Functions for facial feature detection
+static void help(char** argv);
+static void detectFaces(Mat&, vector<Rect_<int> >&, string);
+static void detectEyes(Mat&, vector<Rect_<int> >&, string);
+static void detectNose(Mat&, vector<Rect_<int> >&, string);
+static void detectMouth(Mat&, vector<Rect_<int> >&, string);
+static void detectFacialFeaures(Mat&, const vector<Rect_<int> >, string, string, string);
+
+string input_image_path;
+string face_cascade_path, eye_cascade_path, nose_cascade_path, mouth_cascade_path;
+
+int main(int argc, char** argv)
+{
+    cv::CommandLineParser parser(argc, argv,
+            "{eyes||}{nose||}{mouth||}{help h||}{@image||}{@facexml||}");
+    if (parser.has("help"))
+    {
+        help(argv);
+        return 0;
+    }
+    input_image_path = parser.get<string>("@image");
+    face_cascade_path = parser.get<string>("@facexml");
+    eye_cascade_path = parser.has("eyes") ? parser.get<string>("eyes") : "";
+    nose_cascade_path = parser.has("nose") ? parser.get<string>("nose") : "";
+    mouth_cascade_path = parser.has("mouth") ? parser.get<string>("mouth") : "";
+    if (input_image_path.empty() || face_cascade_path.empty())
+    {
+        cout << "IMAGE or FACE_CASCADE are not specified";
+        return 1;
+    }
+    // Load image and cascade classifier files
+    Mat image;
+    image = imread(samples::findFile(input_image_path));
+
+    // Detect faces and facial features
+    vector<Rect_<int> > faces;
+    detectFaces(image, faces, face_cascade_path);
+    detectFacialFeaures(image, faces, eye_cascade_path, nose_cascade_path, mouth_cascade_path);
+
+    imshow("Result", image);
+
+    waitKey(0);
+    return 0;
+}
+
+static void help(char** argv)
+{
+    cout << "\nThis file demonstrates facial feature points detection using Haarcascade classifiers.\n"
+        "The program detects a face and eyes, nose and mouth inside the face."
+        "The code has been tested on the Japanese Female Facial Expression (JAFFE) database and found"
+        "to give reasonably accurate results. \n";
+
+    cout << "\nUSAGE: " << argv[0] << " [IMAGE] [FACE_CASCADE] [OPTIONS]\n"
+        "IMAGE\n\tPath to the image of a face taken as input.\n"
+        "FACE_CASCSDE\n\t Path to a haarcascade classifier for face detection.\n"
+        "OPTIONS: \nThere are 3 options available which are described in detail. There must be a "
+        "space between the option and it's argument (All three options accept arguments).\n"
+        "\t-eyes=<eyes_cascade> : Specify the haarcascade classifier for eye detection.\n"
+        "\t-nose=<nose_cascade> : Specify the haarcascade classifier for nose detection.\n"
+        "\t-mouth=<mouth-cascade> : Specify the haarcascade classifier for mouth detection.\n";
+
+
+    cout << "EXAMPLE:\n"
+        "(1) " << argv[0] << " image.jpg face.xml -eyes=eyes.xml -mouth=mouth.xml\n"
+        "\tThis will detect the face, eyes and mouth in image.jpg.\n"
+        "(2) " << argv[0] << " image.jpg face.xml -nose=nose.xml\n"
+        "\tThis will detect the face and nose in image.jpg.\n"
+        "(3) " << argv[0] << " image.jpg face.xml\n"
+        "\tThis will detect only the face in image.jpg.\n";
+
+    cout << " \n\nThe classifiers for face and eyes can be downloaded from : "
+        " \nhttps://github.com/opencv/opencv/tree/5.x/data/haarcascades";
+
+    cout << "\n\nThe classifiers for nose and mouth can be downloaded from : "
+        " \nhttps://github.com/opencv/opencv_contrib/tree/5.x/modules/face/data/cascades\n";
+}
+
+static void detectFaces(Mat& img, vector<Rect_<int> >& faces, string cascade_path)
+{
+    CascadeClassifier face_cascade;
+    face_cascade.load(samples::findFile(cascade_path));
+
+    if (!face_cascade.empty())
+        face_cascade.detectMultiScale(img, faces, 1.15, 3, 0|CASCADE_SCALE_IMAGE, Size(30, 30));
+    return;
+}
+
+static void detectFacialFeaures(Mat& img, const vector<Rect_<int> > faces, string eye_cascade,
+        string nose_cascade, string mouth_cascade)
+{
+    for(unsigned int i = 0; i < faces.size(); ++i)
+    {
+        // Mark the bounding box enclosing the face
+        Rect face = faces[i];
+        rectangle(img, Point(face.x, face.y), Point(face.x+face.width, face.y+face.height),
+                Scalar(255, 0, 0), 1, 4);
+
+        // Eyes, nose and mouth will be detected inside the face (region of interest)
+        Mat ROI = img(Rect(face.x, face.y, face.width, face.height));
+
+        // Check if all features (eyes, nose and mouth) are being detected
+        bool is_full_detection = false;
+        if( (!eye_cascade.empty()) && (!nose_cascade.empty()) && (!mouth_cascade.empty()) )
+            is_full_detection = true;
+
+        // Detect eyes if classifier provided by the user
+        if(!eye_cascade.empty())
+        {
+            vector<Rect_<int> > eyes;
+            detectEyes(ROI, eyes, eye_cascade);
+
+            // Mark points corresponding to the centre of the eyes
+            for(unsigned int j = 0; j < eyes.size(); ++j)
+            {
+                Rect e = eyes[j];
+                circle(ROI, Point(e.x+e.width/2, e.y+e.height/2), 3, Scalar(0, 255, 0), -1, 8);
+                /* rectangle(ROI, Point(e.x, e.y), Point(e.x+e.width, e.y+e.height),
+                    Scalar(0, 255, 0), 1, 4); */
+            }
+        }
+
+        // Detect nose if classifier provided by the user
+        double nose_center_height = 0.0;
+        if(!nose_cascade.empty())
+        {
+            vector<Rect_<int> > nose;
+            detectNose(ROI, nose, nose_cascade);
+
+            // Mark points corresponding to the centre (tip) of the nose
+            for(unsigned int j = 0; j < nose.size(); ++j)
+            {
+                Rect n = nose[j];
+                circle(ROI, Point(n.x+n.width/2, n.y+n.height/2), 3, Scalar(0, 255, 0), -1, 8);
+                nose_center_height = (n.y + n.height/2);
+            }
+        }
+
+        // Detect mouth if classifier provided by the user
+        double mouth_center_height = 0.0;
+        if(!mouth_cascade.empty())
+        {
+            vector<Rect_<int> > mouth;
+            detectMouth(ROI, mouth, mouth_cascade);
+
+            for(unsigned int j = 0; j < mouth.size(); ++j)
+            {
+                Rect m = mouth[j];
+                mouth_center_height = (m.y + m.height/2);
+
+                // The mouth should lie below the nose
+                if( (is_full_detection) && (mouth_center_height > nose_center_height) )
+                {
+                    rectangle(ROI, Point(m.x, m.y), Point(m.x+m.width, m.y+m.height), Scalar(0, 255, 0), 1, 4);
+                }
+                else if( (is_full_detection) && (mouth_center_height <= nose_center_height) )
+                    continue;
+                else
+                    rectangle(ROI, Point(m.x, m.y), Point(m.x+m.width, m.y+m.height), Scalar(0, 255, 0), 1, 4);
+            }
+        }
+
+    }
+
+    return;
+}
+
+static void detectEyes(Mat& img, vector<Rect_<int> >& eyes, string cascade_path)
+{
+    CascadeClassifier eyes_cascade;
+    eyes_cascade.load(samples::findFile(cascade_path, !cascade_path.empty()));
+
+    if (!eyes_cascade.empty())
+        eyes_cascade.detectMultiScale(img, eyes, 1.20, 5, 0|CASCADE_SCALE_IMAGE, Size(30, 30));
+    return;
+}
+
+static void detectNose(Mat& img, vector<Rect_<int> >& nose, string cascade_path)
+{
+    CascadeClassifier nose_cascade;
+    nose_cascade.load(samples::findFile(cascade_path, !cascade_path.empty()));
+
+    if (!nose_cascade.empty())
+        nose_cascade.detectMultiScale(img, nose, 1.20, 5, 0|CASCADE_SCALE_IMAGE, Size(30, 30));
+    return;
+}
+
+static void detectMouth(Mat& img, vector<Rect_<int> >& mouth, string cascade_path)
+{
+    CascadeClassifier mouth_cascade;
+    mouth_cascade.load(samples::findFile(cascade_path, !cascade_path.empty()));
+
+    if (!mouth_cascade.empty())
+        mouth_cascade.detectMultiScale(img, mouth, 1.20, 5, 0|CASCADE_SCALE_IMAGE, Size(30, 30));
+    return;
+}

modules/xobjdetect/samples/gpu/cascadeclassifier.cpp

@@ -0,0 +1,316 @@
+// WARNING: this sample is under construction! Use it on your own risk.
+#if defined _MSC_VER && _MSC_VER >= 1400
+#pragma warning(disable : 4100)
+#endif
+
+
+#include <iostream>
+#include <iomanip>
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/cudaobjdetect.hpp"
+#include "opencv2/cudaimgproc.hpp"
+#include "opencv2/cudawarping.hpp"
+
+using namespace std;
+using namespace cv;
+using namespace cv::cuda;
+
+
+static void help()
+{
+    cout << "Usage: ./cascadeclassifier \n\t--cascade <cascade_file>\n\t(<image>|--video <video>|--camera <camera_id>)\n"
+            "Using OpenCV version " << CV_VERSION << endl << endl;
+}
+
+
+static void convertAndResize(const Mat& src, Mat& gray, Mat& resized, double scale)
+{
+    if (src.channels() == 3)
+    {
+        cv::cvtColor( src, gray, COLOR_BGR2GRAY );
+    }
+    else
+    {
+        gray = src;
+    }
+
+    Size sz(cvRound(gray.cols * scale), cvRound(gray.rows * scale));
+
+    if (scale != 1)
+    {
+        cv::resize(gray, resized, sz);
+    }
+    else
+    {
+        resized = gray;
+    }
+}
+
+static void convertAndResize(const GpuMat& src, GpuMat& gray, GpuMat& resized, double scale)
+{
+    if (src.channels() == 3)
+    {
+        cv::cuda::cvtColor( src, gray, COLOR_BGR2GRAY );
+    }
+    else
+    {
+        gray = src;
+    }
+
+    Size sz(cvRound(gray.cols * scale), cvRound(gray.rows * scale));
+
+    if (scale != 1)
+    {
+        cv::cuda::resize(gray, resized, sz);
+    }
+    else
+    {
+        resized = gray;
+    }
+}
+
+
+static void matPrint(Mat &img, int lineOffsY, Scalar fontColor, const string &ss)
+{
+    int fontFace = FONT_HERSHEY_DUPLEX;
+    double fontScale = 0.8;
+    int fontThickness = 2;
+    Size fontSize = cv::getTextSize("T[]", fontFace, fontScale, fontThickness, 0);
+
+    Point org;
+    org.x = 1;
+    org.y = 3 * fontSize.height * (lineOffsY + 1) / 2;
+    putText(img, ss, org, fontFace, fontScale, Scalar(0,0,0), 5*fontThickness/2, 16);
+    putText(img, ss, org, fontFace, fontScale, fontColor, fontThickness, 16);
+}
+
+
+static void displayState(Mat &canvas, bool bHelp, bool bGpu, bool bLargestFace, bool bFilter, double fps)
+{
+    Scalar fontColorRed = Scalar(255,0,0);
+    Scalar fontColorNV  = Scalar(118,185,0);
+
+    ostringstream ss;
+    ss << "FPS = " << setprecision(1) << fixed << fps;
+    matPrint(canvas, 0, fontColorRed, ss.str());
+    ss.str("");
+    ss << "[" << canvas.cols << "x" << canvas.rows << "], " <<
+        (bGpu ? "GPU, " : "CPU, ") <<
+        (bLargestFace ? "OneFace, " : "MultiFace, ") <<
+        (bFilter ? "Filter:ON" : "Filter:OFF");
+    matPrint(canvas, 1, fontColorRed, ss.str());
+
+    // by Anatoly. MacOS fix. ostringstream(const string&) is a private
+    // matPrint(canvas, 2, fontColorNV, ostringstream("Space - switch GPU / CPU"));
+    if (bHelp)
+    {
+        matPrint(canvas, 2, fontColorNV, "Space - switch GPU / CPU");
+        matPrint(canvas, 3, fontColorNV, "M - switch OneFace / MultiFace");
+        matPrint(canvas, 4, fontColorNV, "F - toggle rectangles Filter");
+        matPrint(canvas, 5, fontColorNV, "H - toggle hotkeys help");
+        matPrint(canvas, 6, fontColorNV, "1/Q - increase/decrease scale");
+    }
+    else
+    {
+        matPrint(canvas, 2, fontColorNV, "H - toggle hotkeys help");
+    }
+}
+
+
+int main(int argc, const char *argv[])
+{
+    if (argc == 1)
+    {
+        help();
+        return -1;
+    }
+
+    if (getCudaEnabledDeviceCount() == 0)
+    {
+        return cerr << "No GPU found or the library is compiled without CUDA support" << endl, -1;
+    }
+
+    cv::cuda::printShortCudaDeviceInfo(cv::cuda::getDevice());
+
+    string cascadeName;
+    string inputName;
+    bool isInputImage = false;
+    bool isInputVideo = false;
+    bool isInputCamera = false;
+
+    for (int i = 1; i < argc; ++i)
+    {
+        if (string(argv[i]) == "--cascade")
+            cascadeName = argv[++i];
+        else if (string(argv[i]) == "--video")
+        {
+            inputName = argv[++i];
+            isInputVideo = true;
+        }
+        else if (string(argv[i]) == "--camera")
+        {
+            inputName = argv[++i];
+            isInputCamera = true;
+        }
+        else if (string(argv[i]) == "--help")
+        {
+            help();
+            return -1;
+        }
+        else if (!isInputImage)
+        {
+            inputName = argv[i];
+            isInputImage = true;
+        }
+        else
+        {
+            cout << "Unknown key: " << argv[i] << endl;
+            return -1;
+        }
+    }
+
+    Ptr<cuda::CascadeClassifier> cascade_gpu = cuda::CascadeClassifier::create(cascadeName);
+
+    cv::CascadeClassifier cascade_cpu;
+    if (!cascade_cpu.load(cascadeName))
+    {
+        return cerr << "ERROR: Could not load cascade classifier \"" << cascadeName << "\"" << endl, help(), -1;
+    }
+
+    VideoCapture capture;
+    Mat image;
+
+    if (isInputImage)
+    {
+        image = imread(inputName);
+        CV_Assert(!image.empty());
+    }
+    else if (isInputVideo)
+    {
+        capture.open(inputName);
+        CV_Assert(capture.isOpened());
+    }
+    else
+    {
+        capture.open(atoi(inputName.c_str()));
+        CV_Assert(capture.isOpened());
+    }
+
+    namedWindow("result", 1);
+
+    Mat frame, frame_cpu, gray_cpu, resized_cpu, frameDisp;
+    vector<Rect> faces;
+
+    GpuMat frame_gpu, gray_gpu, resized_gpu, facesBuf_gpu;
+
+    /* parameters */
+    bool useGPU = true;
+    double scaleFactor = 1.0;
+    bool findLargestObject = false;
+    bool filterRects = true;
+    bool helpScreen = false;
+
+    for (;;)
+    {
+        if (isInputCamera || isInputVideo)
+        {
+            capture >> frame;
+            if (frame.empty())
+            {
+                break;
+            }
+        }
+
+        (image.empty() ? frame : image).copyTo(frame_cpu);
+        frame_gpu.upload(image.empty() ? frame : image);
+
+        convertAndResize(frame_gpu, gray_gpu, resized_gpu, scaleFactor);
+        convertAndResize(frame_cpu, gray_cpu, resized_cpu, scaleFactor);
+
+        TickMeter tm;
+        tm.start();
+
+        if (useGPU)
+        {
+            cascade_gpu->setFindLargestObject(findLargestObject);
+            cascade_gpu->setScaleFactor(1.2);
+            cascade_gpu->setMinNeighbors((filterRects || findLargestObject) ? 4 : 0);
+
+            cascade_gpu->detectMultiScale(resized_gpu, facesBuf_gpu);
+            cascade_gpu->convert(facesBuf_gpu, faces);
+        }
+        else
+        {
+            Size minSize = cascade_gpu->getClassifierSize();
+            cascade_cpu.detectMultiScale(resized_cpu, faces, 1.2,
+                                         (filterRects || findLargestObject) ? 4 : 0,
+                                         (findLargestObject ? CASCADE_FIND_BIGGEST_OBJECT : 0)
+                                            | CASCADE_SCALE_IMAGE,
+                                         minSize);
+        }
+
+        for (size_t i = 0; i < faces.size(); ++i)
+        {
+            rectangle(resized_cpu, faces[i], Scalar(255));
+        }
+
+        tm.stop();
+        double detectionTime = tm.getTimeMilli();
+        double fps = 1000 / detectionTime;
+
+        //print detections to console
+        cout << setfill(' ') << setprecision(2);
+        cout << setw(6) << fixed << fps << " FPS, " << faces.size() << " det";
+        if ((filterRects || findLargestObject) && !faces.empty())
+        {
+            for (size_t i = 0; i < faces.size(); ++i)
+            {
+                cout << ", [" << setw(4) << faces[i].x
+                     << ", " << setw(4) << faces[i].y
+                     << ", " << setw(4) << faces[i].width
+                     << ", " << setw(4) << faces[i].height << "]";
+            }
+        }
+        cout << endl;
+
+        cv::cvtColor(resized_cpu, frameDisp, COLOR_GRAY2BGR);
+        displayState(frameDisp, helpScreen, useGPU, findLargestObject, filterRects, fps);
+        imshow("result", frameDisp);
+
+        char key = (char)waitKey(5);
+        if (key == 27)
+        {
+            break;
+        }
+
+        switch (key)
+        {
+        case ' ':
+            useGPU = !useGPU;
+            break;
+        case 'm':
+        case 'M':
+            findLargestObject = !findLargestObject;
+            break;
+        case 'f':
+        case 'F':
+            filterRects = !filterRects;
+            break;
+        case '1':
+            scaleFactor *= 1.05;
+            break;
+        case 'q':
+        case 'Q':
+            scaleFactor /= 1.05;
+            break;
+        case 'h':
+        case 'H':
+            helpScreen = !helpScreen;
+            break;
+        }
+    }
+
+    return 0;
+}

modules/xobjdetect/samples/gpu/hog.cpp

@@ -0,0 +1,552 @@
+#include <iostream>
+#include <fstream>
+#include <string>
+#include <sstream>
+#include <iomanip>
+#include <stdexcept>
+#include <opencv2/core/utility.hpp>
+#include "opencv2/cudaobjdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/imgproc.hpp"
+
+using namespace std;
+using namespace cv;
+
+bool help_showed = false;
+
+class Args
+{
+public:
+    Args();
+    static Args read(int argc, char** argv);
+
+    string src;
+    bool src_is_folder;
+    bool src_is_video;
+    bool src_is_camera;
+    int camera_id;
+
+    bool svm_load;
+    string svm;
+
+    bool write_video;
+    string dst_video;
+    double dst_video_fps;
+
+    bool make_gray;
+
+    bool resize_src;
+    int width, height;
+
+    double scale;
+    int nlevels;
+    int gr_threshold;
+
+    double hit_threshold;
+    bool hit_threshold_auto;
+
+    int win_width;
+    int win_stride_width, win_stride_height;
+    int block_width;
+    int block_stride_width, block_stride_height;
+    int cell_width;
+    int nbins;
+
+    bool gamma_corr;
+};
+
+
+class App
+{
+public:
+    App(const Args& s);
+    void run();
+
+    void handleKey(char key);
+
+    void hogWorkBegin();
+    void hogWorkEnd();
+    string hogWorkFps() const;
+
+    void workBegin();
+    void workEnd();
+    string workFps() const;
+
+    string message() const;
+
+private:
+    App operator=(App&);
+
+    Args args;
+    bool running;
+
+    bool use_gpu;
+    bool make_gray;
+    double scale;
+    int gr_threshold;
+    int nlevels;
+    double hit_threshold;
+    bool gamma_corr;
+
+    int64 hog_work_begin;
+    double hog_work_fps;
+
+    int64 work_begin;
+    double work_fps;
+};
+
+static void printHelp()
+{
+    cout << "Histogram of Oriented Gradients descriptor and detector sample.\n"
+         << "\nUsage: hog\n"
+         << "  (<image>|--video <vide>|--camera <camera_id>) # frames source\n"
+         << "  or"
+         << "  (--folder <folder_path>) # load images from folder\n"
+         << "  [--svm <file> # load svm file"
+         << "  [--make_gray <true/false>] # convert image to gray one or not\n"
+         << "  [--resize_src <true/false>] # do resize of the source image or not\n"
+         << "  [--width <int>] # resized image width\n"
+         << "  [--height <int>] # resized image height\n"
+         << "  [--hit_threshold <double>] # classifying plane distance threshold (0.0 usually)\n"
+         << "  [--scale <double>] # HOG window scale factor\n"
+         << "  [--nlevels <int>] # max number of HOG window scales\n"
+         << "  [--win_width <int>] # width of the window\n"
+         << "  [--win_stride_width <int>] # distance by OX axis between neighbour wins\n"
+         << "  [--win_stride_height <int>] # distance by OY axis between neighbour wins\n"
+         << "  [--block_width <int>] # width of the block\n"
+         << "  [--block_stride_width <int>] # distance by 0X axis between neighbour blocks\n"
+         << "  [--block_stride_height <int>] # distance by 0Y axis between neighbour blocks\n"
+         << "  [--cell_width <int>] # width of the cell\n"
+         << "  [--nbins <int>] # number of bins\n"
+         << "  [--gr_threshold <int>] # merging similar rects constant\n"
+         << "  [--gamma_correct <int>] # do gamma correction or not\n"
+         << "  [--write_video <bool>] # write video or not\n"
+         << "  [--dst_video <path>] # output video path\n"
+         << "  [--dst_video_fps <double>] # output video fps\n";
+    help_showed = true;
+}
+
+int main(int argc, char** argv)
+{
+    try
+    {
+        Args args;
+        if (argc < 2)
+        {
+            printHelp();
+            args.camera_id = 0;
+            args.src_is_camera = true;
+        }
+        else
+        {
+            args = Args::read(argc, argv);
+            if (help_showed)
+                return -1;
+        }
+        App app(args);
+        app.run();
+    }
+    catch (const Exception& e) { return cout << "error: "  << e.what() << endl, 1; }
+    catch (const exception& e) { return cout << "error: "  << e.what() << endl, 1; }
+    catch(...) { return cout << "unknown exception" << endl, 1; }
+    return 0;
+}
+
+
+Args::Args()
+{
+    src_is_video = false;
+    src_is_camera = false;
+    src_is_folder = false;
+    svm_load = false;
+    camera_id = 0;
+
+    write_video = false;
+    dst_video_fps = 24.;
+
+    make_gray = false;
+
+    resize_src = false;
+    width = 640;
+    height = 480;
+
+    scale = 1.05;
+    nlevels = 13;
+    gr_threshold = 8;
+    hit_threshold = 1.4;
+    hit_threshold_auto = true;
+
+    win_width = 48;
+    win_stride_width = 8;
+    win_stride_height = 8;
+    block_width = 16;
+    block_stride_width = 8;
+    block_stride_height = 8;
+    cell_width = 8;
+    nbins = 9;
+
+    gamma_corr = true;
+}
+
+
+Args Args::read(int argc, char** argv)
+{
+    Args args;
+    for (int i = 1; i < argc; i++)
+    {
+        if (string(argv[i]) == "--make_gray") args.make_gray = (string(argv[++i]) == "true");
+        else if (string(argv[i]) == "--resize_src") args.resize_src = (string(argv[++i]) == "true");
+        else if (string(argv[i]) == "--width") args.width = atoi(argv[++i]);
+        else if (string(argv[i]) == "--height") args.height = atoi(argv[++i]);
+        else if (string(argv[i]) == "--hit_threshold")
+        {
+            args.hit_threshold = atof(argv[++i]);
+            args.hit_threshold_auto = false;
+        }
+        else if (string(argv[i]) == "--scale") args.scale = atof(argv[++i]);
+        else if (string(argv[i]) == "--nlevels") args.nlevels = atoi(argv[++i]);
+        else if (string(argv[i]) == "--win_width") args.win_width = atoi(argv[++i]);
+        else if (string(argv[i]) == "--win_stride_width") args.win_stride_width = atoi(argv[++i]);
+        else if (string(argv[i]) == "--win_stride_height") args.win_stride_height = atoi(argv[++i]);
+        else if (string(argv[i]) == "--block_width") args.block_width = atoi(argv[++i]);
+        else if (string(argv[i]) == "--block_stride_width") args.block_stride_width = atoi(argv[++i]);
+        else if (string(argv[i]) == "--block_stride_height") args.block_stride_height = atoi(argv[++i]);
+        else if (string(argv[i]) == "--cell_width") args.cell_width = atoi(argv[++i]);
+        else if (string(argv[i]) == "--nbins") args.nbins = atoi(argv[++i]);
+        else if (string(argv[i]) == "--gr_threshold") args.gr_threshold = atoi(argv[++i]);
+        else if (string(argv[i]) == "--gamma_correct") args.gamma_corr = (string(argv[++i]) == "true");
+        else if (string(argv[i]) == "--write_video") args.write_video = (string(argv[++i]) == "true");
+        else if (string(argv[i]) == "--dst_video") args.dst_video = argv[++i];
+        else if (string(argv[i]) == "--dst_video_fps") args.dst_video_fps = atof(argv[++i]);
+        else if (string(argv[i]) == "--help") printHelp();
+        else if (string(argv[i]) == "--video") { args.src = argv[++i]; args.src_is_video = true; }
+        else if (string(argv[i]) == "--camera") { args.camera_id = atoi(argv[++i]); args.src_is_camera = true; }
+        else if (string(argv[i]) == "--folder") { args.src = argv[++i]; args.src_is_folder = true;}
+        else if (string(argv[i]) == "--svm") { args.svm = argv[++i]; args.svm_load = true;}
+        else if (args.src.empty()) args.src = argv[i];
+        else throw runtime_error((string("unknown key: ") + argv[i]));
+    }
+    return args;
+}
+
+
+App::App(const Args& s)
+{
+    cv::cuda::printShortCudaDeviceInfo(cv::cuda::getDevice());
+
+    args = s;
+    cout << "\nControls:\n"
+         << "\tESC - exit\n"
+         << "\tm - change mode GPU <-> CPU\n"
+         << "\tg - convert image to gray or not\n"
+         << "\t1/q - increase/decrease HOG scale\n"
+         << "\t2/w - increase/decrease levels count\n"
+         << "\t3/e - increase/decrease HOG group threshold\n"
+         << "\t4/r - increase/decrease hit threshold\n"
+         << endl;
+
+    use_gpu = true;
+    make_gray = args.make_gray;
+    scale = args.scale;
+    gr_threshold = args.gr_threshold;
+    nlevels = args.nlevels;
+
+    if (args.hit_threshold_auto)
+        args.hit_threshold = args.win_width == 48 ? 1.4 : 0.;
+    hit_threshold = args.hit_threshold;
+
+    gamma_corr = args.gamma_corr;
+
+    cout << "Scale: " << scale << endl;
+    if (args.resize_src)
+        cout << "Resized source: (" << args.width << ", " << args.height << ")\n";
+    cout << "Group threshold: " << gr_threshold << endl;
+    cout << "Levels number: " << nlevels << endl;
+    cout << "Win size: (" << args.win_width << ", " << args.win_width*2 << ")\n";
+    cout << "Win stride: (" << args.win_stride_width << ", " << args.win_stride_height << ")\n";
+    cout << "Block size: (" << args.block_width << ", " << args.block_width << ")\n";
+    cout << "Block stride: (" << args.block_stride_width << ", " << args.block_stride_height << ")\n";
+    cout << "Cell size: (" << args.cell_width << ", " << args.cell_width << ")\n";
+    cout << "Bins number: " << args.nbins << endl;
+    cout << "Hit threshold: " << hit_threshold << endl;
+    cout << "Gamma correction: " << gamma_corr << endl;
+    cout << endl;
+}
+
+
+void App::run()
+{
+    running = true;
+    cv::VideoWriter video_writer;
+
+    Size win_stride(args.win_stride_width, args.win_stride_height);
+    Size win_size(args.win_width, args.win_width * 2);
+    Size block_size(args.block_width, args.block_width);
+    Size block_stride(args.block_stride_width, args.block_stride_height);
+    Size cell_size(args.cell_width, args.cell_width);
+
+    cv::Ptr<cv::cuda::HOG> gpu_hog = cv::cuda::HOG::create(win_size, block_size, block_stride, cell_size, args.nbins);
+    cv::HOGDescriptor cpu_hog(win_size, block_size, block_stride, cell_size, args.nbins);
+
+    if(args.svm_load) {
+        std::vector<float> svm_model;
+        const std::string model_file_name = args.svm;
+        FileStorage ifs(model_file_name, FileStorage::READ);
+        if (ifs.isOpened()) {
+            ifs["svm_detector"] >> svm_model;
+        } else {
+            const std::string what =
+                    "could not load model for hog classifier from file: "
+                    + model_file_name;
+            throw std::runtime_error(what);
+        }
+
+        // check if the variables are initialized
+        if (svm_model.empty()) {
+            const std::string what =
+                    "HoG classifier: svm model could not be loaded from file"
+                    + model_file_name;
+            throw std::runtime_error(what);
+        }
+
+        gpu_hog->setSVMDetector(svm_model);
+        cpu_hog.setSVMDetector(svm_model);
+    } else {
+        // Create HOG descriptors and detectors here
+        Mat detector = gpu_hog->getDefaultPeopleDetector();
+
+        gpu_hog->setSVMDetector(detector);
+        cpu_hog.setSVMDetector(detector);
+    }
+
+    cout << "gpusvmDescriptorSize : " << gpu_hog->getDescriptorSize()
+         << endl;
+    cout << "cpusvmDescriptorSize : " << cpu_hog.getDescriptorSize()
+         << endl;
+
+    while (running)
+    {
+        VideoCapture vc;
+        Mat frame;
+        vector<String> filenames;
+
+        unsigned int count = 1;
+
+        if (args.src_is_video)
+        {
+            vc.open(args.src.c_str());
+            if (!vc.isOpened())
+                throw runtime_error(string("can't open video file: " + args.src));
+            vc >> frame;
+        }
+        else if (args.src_is_folder) {
+            String folder = args.src;
+            cout << folder << endl;
+            glob(folder, filenames);
+            frame = imread(filenames[count]);	// 0 --> .gitignore
+            if (!frame.data)
+                cerr << "Problem loading image from folder!!!" << endl;
+        }
+        else if (args.src_is_camera)
+        {
+            vc.open(args.camera_id);
+            if (!vc.isOpened())
+            {
+                stringstream msg;
+                msg << "can't open camera: " << args.camera_id;
+                throw runtime_error(msg.str());
+            }
+            vc >> frame;
+        }
+        else
+        {
+            frame = imread(args.src);
+            if (frame.empty())
+                throw runtime_error(string("can't open image file: " + args.src));
+        }
+
+        Mat img_aux, img, img_to_show;
+        cuda::GpuMat gpu_img;
+
+        // Iterate over all frames
+        while (running && !frame.empty())
+        {
+            workBegin();
+
+            // Change format of the image
+            if (make_gray) cvtColor(frame, img_aux, COLOR_BGR2GRAY);
+            else if (use_gpu) cvtColor(frame, img_aux, COLOR_BGR2BGRA);
+            else frame.copyTo(img_aux);
+
+            // Resize image
+            if (args.resize_src) resize(img_aux, img, Size(args.width, args.height));
+            else img = img_aux;
+            img_to_show = img;
+
+            vector<Rect> found;
+
+            // Perform HOG classification
+            hogWorkBegin();
+            if (use_gpu)
+            {
+                gpu_img.upload(img);
+                gpu_hog->setNumLevels(nlevels);
+                gpu_hog->setHitThreshold(hit_threshold);
+                gpu_hog->setWinStride(win_stride);
+                gpu_hog->setScaleFactor(scale);
+                gpu_hog->setGroupThreshold(gr_threshold);
+                gpu_hog->detectMultiScale(gpu_img, found);
+            }
+            else
+            {
+                cpu_hog.nlevels = nlevels;
+                cpu_hog.detectMultiScale(img, found, hit_threshold, win_stride,
+                                         Size(0, 0), scale, gr_threshold);
+            }
+            hogWorkEnd();
+
+            // Draw positive classified windows
+            for (size_t i = 0; i < found.size(); i++)
+            {
+                Rect r = found[i];
+                rectangle(img_to_show, r.tl(), r.br(), Scalar(0, 255, 0), 3);
+            }
+
+            if (use_gpu)
+                putText(img_to_show, "Mode: GPU", Point(5, 25), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
+            else
+                putText(img_to_show, "Mode: CPU", Point(5, 25), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
+            putText(img_to_show, "FPS HOG: " + hogWorkFps(), Point(5, 65), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
+            putText(img_to_show, "FPS total: " + workFps(), Point(5, 105), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
+            imshow("opencv_gpu_hog", img_to_show);
+
+            if (args.src_is_video || args.src_is_camera) vc >> frame;
+            if (args.src_is_folder) {
+                count++;
+                if (count < filenames.size()) {
+                    frame = imread(filenames[count]);
+                } else {
+                    Mat empty;
+                    frame = empty;
+                }
+            }
+
+            workEnd();
+
+            if (args.write_video)
+            {
+                if (!video_writer.isOpened())
+                {
+                    video_writer.open(args.dst_video, VideoWriter::fourcc('x','v','i','d'), args.dst_video_fps,
+                                      img_to_show.size(), true);
+                    if (!video_writer.isOpened())
+                        throw std::runtime_error("can't create video writer");
+                }
+
+                if (make_gray) cvtColor(img_to_show, img, COLOR_GRAY2BGR);
+                else cvtColor(img_to_show, img, COLOR_BGRA2BGR);
+
+                video_writer << img;
+            }
+
+            handleKey((char)waitKey(3));
+        }
+    }
+}
+
+
+void App::handleKey(char key)
+{
+    switch (key)
+    {
+    case 27:
+        running = false;
+        break;
+    case 'm':
+    case 'M':
+        use_gpu = !use_gpu;
+        cout << "Switched to " << (use_gpu ? "CUDA" : "CPU") << " mode\n";
+        break;
+    case 'g':
+    case 'G':
+        make_gray = !make_gray;
+        cout << "Convert image to gray: " << (make_gray ? "YES" : "NO") << endl;
+        break;
+    case '1':
+        scale *= 1.05;
+        cout << "Scale: " << scale << endl;
+        break;
+    case 'q':
+    case 'Q':
+        scale /= 1.05;
+        cout << "Scale: " << scale << endl;
+        break;
+    case '2':
+        nlevels++;
+        cout << "Levels number: " << nlevels << endl;
+        break;
+    case 'w':
+    case 'W':
+        nlevels = max(nlevels - 1, 1);
+        cout << "Levels number: " << nlevels << endl;
+        break;
+    case '3':
+        gr_threshold++;
+        cout << "Group threshold: " << gr_threshold << endl;
+        break;
+    case 'e':
+    case 'E':
+        gr_threshold = max(0, gr_threshold - 1);
+        cout << "Group threshold: " << gr_threshold << endl;
+        break;
+    case '4':
+        hit_threshold+=0.25;
+        cout << "Hit threshold: " << hit_threshold << endl;
+        break;
+    case 'r':
+    case 'R':
+        hit_threshold = max(0.0, hit_threshold - 0.25);
+        cout << "Hit threshold: " << hit_threshold << endl;
+        break;
+    case 'c':
+    case 'C':
+        gamma_corr = !gamma_corr;
+        cout << "Gamma correction: " << gamma_corr << endl;
+        break;
+    }
+}
+
+
+inline void App::hogWorkBegin() { hog_work_begin = getTickCount(); }
+
+inline void App::hogWorkEnd()
+{
+    int64 delta = getTickCount() - hog_work_begin;
+    double freq = getTickFrequency();
+    hog_work_fps = freq / delta;
+}
+
+inline string App::hogWorkFps() const
+{
+    stringstream ss;
+    ss << hog_work_fps;
+    return ss.str();
+}
+
+
+inline void App::workBegin() { work_begin = getTickCount(); }
+
+inline void App::workEnd()
+{
+    int64 delta = getTickCount() - work_begin;
+    double freq = getTickFrequency();
+    work_fps = freq / delta;
+}
+
+inline string App::workFps() const
+{
+    stringstream ss;
+    ss << work_fps;
+    return ss.str();
+}

modules/xobjdetect/samples/hog_tapi.cpp

@@ -0,0 +1,352 @@
+#include <iostream>
+#include <fstream>
+#include <string>
+#include <sstream>
+#include <iomanip>
+#include <stdexcept>
+#include <opencv2/core/ocl.hpp>
+#include <opencv2/core/utility.hpp>
+#include "opencv2/imgcodecs.hpp"
+#include <opencv2/videoio.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/xobjdetect.hpp>
+#include <opencv2/imgproc.hpp>
+
+using namespace std;
+using namespace cv;
+
+class App
+{
+public:
+    App(CommandLineParser& cmd);
+    void run();
+    void handleKey(char key);
+    void hogWorkBegin();
+    void hogWorkEnd();
+    string hogWorkFps() const;
+    void workBegin();
+    void workEnd();
+    string workFps() const;
+private:
+    App operator=(App&);
+
+    //Args args;
+    bool running;
+    bool make_gray;
+    double scale;
+    double resize_scale;
+    int win_width;
+    int win_stride_width, win_stride_height;
+    int gr_threshold;
+    int nlevels;
+    double hit_threshold;
+    bool gamma_corr;
+
+    int64 hog_work_begin;
+    double hog_work_fps;
+    int64 work_begin;
+    double work_fps;
+
+    string img_source;
+    string vdo_source;
+    string output;
+    int camera_id;
+    bool write_once;
+};
+
+int main(int argc, char** argv)
+{
+    const char* keys =
+        "{ h help      |                | print help message }"
+        "{ i input     |                | specify input image}"
+        "{ c camera    | -1             | enable camera capturing }"
+        "{ v video     | vtest.avi | use video as input }"
+        "{ g gray      |                | convert image to gray one or not}"
+        "{ s scale     | 1.0            | resize the image before detect}"
+        "{ o output    |   output.avi   | specify output path when input is images}";
+    CommandLineParser cmd(argc, argv, keys);
+    if (cmd.has("help"))
+    {
+        cmd.printMessage();
+        return EXIT_SUCCESS;
+    }
+
+    App app(cmd);
+    try
+    {
+        app.run();
+    }
+    catch (const Exception& e)
+    {
+        return cout << "error: "  << e.what() << endl, 1;
+    }
+    catch (const exception& e)
+    {
+        return cout << "error: "  << e.what() << endl, 1;
+    }
+    catch(...)
+    {
+        return cout << "unknown exception" << endl, 1;
+    }
+    return EXIT_SUCCESS;
+}
+
+App::App(CommandLineParser& cmd)
+{
+    cout << "\nControls:\n"
+         << "\tESC - exit\n"
+         << "\tm - change mode GPU <-> CPU\n"
+         << "\tg - convert image to gray or not\n"
+         << "\to - save output image once, or switch on/off video save\n"
+         << "\t1/q - increase/decrease HOG scale\n"
+         << "\t2/w - increase/decrease levels count\n"
+         << "\t3/e - increase/decrease HOG group threshold\n"
+         << "\t4/r - increase/decrease hit threshold\n"
+         << endl;
+
+    make_gray = cmd.has("gray");
+    resize_scale = cmd.get<double>("s");
+    vdo_source = samples::findFileOrKeep(cmd.get<string>("v"));
+    img_source = cmd.get<string>("i");
+    output = cmd.get<string>("o");
+    camera_id = cmd.get<int>("c");
+
+    win_width = 48;
+    win_stride_width = 8;
+    win_stride_height = 8;
+    gr_threshold = 8;
+    nlevels = 13;
+    hit_threshold = 1.4;
+    scale = 1.05;
+    gamma_corr = true;
+    write_once = false;
+
+    cout << "Group threshold: " << gr_threshold << endl;
+    cout << "Levels number: " << nlevels << endl;
+    cout << "Win width: " << win_width << endl;
+    cout << "Win stride: (" << win_stride_width << ", " << win_stride_height << ")\n";
+    cout << "Hit threshold: " << hit_threshold << endl;
+    cout << "Gamma correction: " << gamma_corr << endl;
+    cout << endl;
+}
+
+void App::run()
+{
+    running = true;
+    VideoWriter video_writer;
+
+    Size win_size(win_width, win_width * 2);
+    Size win_stride(win_stride_width, win_stride_height);
+
+    // Create HOG descriptors and detectors here
+
+    HOGDescriptor hog(win_size, Size(16, 16), Size(8, 8), Size(8, 8), 9, 1, -1,
+                          HOGDescriptor::L2Hys, 0.2, gamma_corr, cv::HOGDescriptor::DEFAULT_NLEVELS);
+    hog.setSVMDetector( HOGDescriptor::getDaimlerPeopleDetector() );
+
+    while (running)
+    {
+        VideoCapture vc;
+        UMat frame;
+
+        if (vdo_source!="")
+        {
+            vc.open(vdo_source.c_str());
+            if (!vc.isOpened())
+                throw runtime_error(string("can't open video file: " + vdo_source));
+            vc >> frame;
+        }
+        else if (camera_id != -1)
+        {
+            vc.open(camera_id);
+            if (!vc.isOpened())
+            {
+                stringstream msg;
+                msg << "can't open camera: " << camera_id;
+                throw runtime_error(msg.str());
+            }
+            vc >> frame;
+        }
+        else
+        {
+            imread(img_source).copyTo(frame);
+            if (frame.empty())
+                throw runtime_error(string("can't open image file: " + img_source));
+        }
+
+        UMat img_aux, img, img_to_show;
+
+        // Iterate over all frames
+        while (running && !frame.empty())
+        {
+            workBegin();
+
+            // Change format of the image
+            if (make_gray) cvtColor(frame, img_aux, COLOR_BGR2GRAY );
+            else frame.copyTo(img_aux);
+
+            // Resize image
+            if (abs(scale-1.0)>0.001)
+            {
+                Size sz((int)((double)img_aux.cols/resize_scale), (int)((double)img_aux.rows/resize_scale));
+                resize(img_aux, img, sz, 0, 0, INTER_LINEAR_EXACT);
+            }
+            else img = img_aux;
+            img.copyTo(img_to_show);
+            hog.nlevels = nlevels;
+            vector<Rect> found;
+
+            // Perform HOG classification
+            hogWorkBegin();
+
+            hog.detectMultiScale(img, found, hit_threshold, win_stride,
+                    Size(0, 0), scale, gr_threshold);
+            hogWorkEnd();
+
+
+            // Draw positive classified windows
+            for (size_t i = 0; i < found.size(); i++)
+            {
+                rectangle(img_to_show, found[i], Scalar(0, 255, 0), 3);
+            }
+
+            putText(img_to_show, ocl::useOpenCL() ? "Mode: OpenCL"  : "Mode: CPU", Point(5, 25), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
+            putText(img_to_show, "FPS (HOG only): " + hogWorkFps(), Point(5, 65), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
+            putText(img_to_show, "FPS (total): " + workFps(), Point(5, 105), FONT_HERSHEY_SIMPLEX, 1., Scalar(255, 100, 0), 2);
+            imshow("opencv_hog", img_to_show);
+            if (vdo_source!="" || camera_id!=-1) vc >> frame;
+
+            workEnd();
+
+            if (output!="" && write_once)
+            {
+                if (img_source!="")     // write image
+                {
+                    write_once = false;
+                    imwrite(output, img_to_show);
+                }
+                else                    //write video
+                {
+                    if (!video_writer.isOpened())
+                    {
+                        video_writer.open(output, VideoWriter::fourcc('x','v','i','d'), 24,
+                                          img_to_show.size(), true);
+                        if (!video_writer.isOpened())
+                            throw std::runtime_error("can't create video writer");
+                    }
+
+                    if (make_gray) cvtColor(img_to_show, img, COLOR_GRAY2BGR);
+                    else cvtColor(img_to_show, img, COLOR_BGRA2BGR);
+
+                    video_writer << img;
+                }
+            }
+
+            handleKey((char)waitKey(3));
+        }
+    }
+}
+
+void App::handleKey(char key)
+{
+    switch (key)
+    {
+    case 27:
+        running = false;
+        break;
+    case 'm':
+    case 'M':
+        ocl::setUseOpenCL(!cv::ocl::useOpenCL());
+        cout << "Switched to " << (ocl::useOpenCL() ? "OpenCL enabled" : "CPU") << " mode\n";
+        break;
+    case 'g':
+    case 'G':
+        make_gray = !make_gray;
+        cout << "Convert image to gray: " << (make_gray ? "YES" : "NO") << endl;
+        break;
+    case '1':
+        scale *= 1.05;
+        cout << "Scale: " << scale << endl;
+        break;
+    case 'q':
+    case 'Q':
+        scale /= 1.05;
+        cout << "Scale: " << scale << endl;
+        break;
+    case '2':
+        nlevels++;
+        cout << "Levels number: " << nlevels << endl;
+        break;
+    case 'w':
+    case 'W':
+        nlevels = max(nlevels - 1, 1);
+        cout << "Levels number: " << nlevels << endl;
+        break;
+    case '3':
+        gr_threshold++;
+        cout << "Group threshold: " << gr_threshold << endl;
+        break;
+    case 'e':
+    case 'E':
+        gr_threshold = max(0, gr_threshold - 1);
+        cout << "Group threshold: " << gr_threshold << endl;
+        break;
+    case '4':
+        hit_threshold+=0.25;
+        cout << "Hit threshold: " << hit_threshold << endl;
+        break;
+    case 'r':
+    case 'R':
+        hit_threshold = max(0.0, hit_threshold - 0.25);
+        cout << "Hit threshold: " << hit_threshold << endl;
+        break;
+    case 'c':
+    case 'C':
+        gamma_corr = !gamma_corr;
+        cout << "Gamma correction: " << gamma_corr << endl;
+        break;
+    case 'o':
+    case 'O':
+        write_once = !write_once;
+        break;
+    }
+}
+
+
+inline void App::hogWorkBegin()
+{
+    hog_work_begin = getTickCount();
+}
+
+inline void App::hogWorkEnd()
+{
+    int64 delta = getTickCount() - hog_work_begin;
+    double freq = getTickFrequency();
+    hog_work_fps = freq / delta;
+}
+
+inline string App::hogWorkFps() const
+{
+    stringstream ss;
+    ss << hog_work_fps;
+    return ss.str();
+}
+
+inline void App::workBegin()
+{
+    work_begin = getTickCount();
+}
+
+inline void App::workEnd()
+{
+    int64 delta = getTickCount() - work_begin;
+    double freq = getTickFrequency();
+    work_fps = freq / delta;
+}
+
+inline string App::workFps() const
+{
+    stringstream ss;
+    ss << work_fps;
+    return ss.str();
+}

modules/xobjdetect/samples/java/tutorial_code/objectDetection/cascade_classifier/ObjectDetectionDemo.java

@@ -0,0 +1,98 @@
+import java.util.List;
+
+import org.opencv.core.Core;
+import org.opencv.core.Mat;
+import org.opencv.core.MatOfRect;
+import org.opencv.core.Point;
+import org.opencv.core.Rect;
+import org.opencv.core.Scalar;
+import org.opencv.core.Size;
+import org.opencv.highgui.HighGui;
+import org.opencv.imgproc.Imgproc;
+import org.opencv.xobjdetect.CascadeClassifier;
+import org.opencv.videoio.VideoCapture;
+
+class ObjectDetection {
+    public void detectAndDisplay(Mat frame, CascadeClassifier faceCascade, CascadeClassifier eyesCascade) {
+        Mat frameGray = new Mat();
+        Imgproc.cvtColor(frame, frameGray, Imgproc.COLOR_BGR2GRAY);
+        Imgproc.equalizeHist(frameGray, frameGray);
+
+        // -- Detect faces
+        MatOfRect faces = new MatOfRect();
+        faceCascade.detectMultiScale(frameGray, faces);
+
+        List<Rect> listOfFaces = faces.toList();
+        for (Rect face : listOfFaces) {
+            Point center = new Point(face.x + face.width / 2, face.y + face.height / 2);
+            Imgproc.ellipse(frame, center, new Size(face.width / 2, face.height / 2), 0, 0, 360,
+                    new Scalar(255, 0, 255));
+
+            Mat faceROI = frameGray.submat(face);
+
+            // -- In each face, detect eyes
+            MatOfRect eyes = new MatOfRect();
+            eyesCascade.detectMultiScale(faceROI, eyes);
+
+            List<Rect> listOfEyes = eyes.toList();
+            for (Rect eye : listOfEyes) {
+                Point eyeCenter = new Point(face.x + eye.x + eye.width / 2, face.y + eye.y + eye.height / 2);
+                int radius = (int) Math.round((eye.width + eye.height) * 0.25);
+                Imgproc.circle(frame, eyeCenter, radius, new Scalar(255, 0, 0), 4);
+            }
+        }
+
+        //-- Show what you got
+        HighGui.imshow("Capture - Face detection", frame );
+    }
+
+    public void run(String[] args) {
+        String filenameFaceCascade = args.length > 2 ? args[0] : "../../data/haarcascades/haarcascade_frontalface_alt.xml";
+        String filenameEyesCascade = args.length > 2 ? args[1] : "../../data/haarcascades/haarcascade_eye_tree_eyeglasses.xml";
+        int cameraDevice = args.length > 2 ? Integer.parseInt(args[2]) : 0;
+
+        CascadeClassifier faceCascade = new CascadeClassifier();
+        CascadeClassifier eyesCascade = new CascadeClassifier();
+
+        if (!faceCascade.load(filenameFaceCascade)) {
+            System.err.println("--(!)Error loading face cascade: " + filenameFaceCascade);
+            System.exit(0);
+        }
+        if (!eyesCascade.load(filenameEyesCascade)) {
+            System.err.println("--(!)Error loading eyes cascade: " + filenameEyesCascade);
+            System.exit(0);
+        }
+
+        VideoCapture capture = new VideoCapture(cameraDevice);
+        if (!capture.isOpened()) {
+            System.err.println("--(!)Error opening video capture");
+            System.exit(0);
+        }
+
+        Mat frame = new Mat();
+        while (capture.read(frame)) {
+            if (frame.empty()) {
+                System.err.println("--(!) No captured frame -- Break!");
+                break;
+            }
+
+            //-- 3. Apply the classifier to the frame
+            detectAndDisplay(frame, faceCascade, eyesCascade);
+
+            if (HighGui.waitKey(10) == 27) {
+                break;// escape
+            }
+        }
+
+        System.exit(0);
+    }
+}
+
+public class ObjectDetectionDemo {
+    public static void main(String[] args) {
+        // Load the native OpenCV library
+        System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
+
+        new ObjectDetection().run(args);
+    }
+}

modules/xobjdetect/samples/objectDetection.cpp

@@ -0,0 +1,107 @@
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/videoio.hpp"
+#include <iostream>
+
+using namespace std;
+using namespace cv;
+
+/** Function Headers */
+void detectAndDisplay( Mat frame );
+
+/** Global variables */
+CascadeClassifier face_cascade;
+CascadeClassifier eyes_cascade;
+
+/** @function main */
+int main( int argc, const char** argv )
+{
+    CommandLineParser parser(argc, argv,
+                             "{help h||}"
+                             "{face_cascade|data/haarcascades/haarcascade_frontalface_alt.xml|Path to face cascade.}"
+                             "{eyes_cascade|data/haarcascades/haarcascade_eye_tree_eyeglasses.xml|Path to eyes cascade.}"
+                             "{camera|0|Camera device number.}");
+
+    parser.about( "\nThis program demonstrates using the cv::CascadeClassifier class to detect objects (Face + eyes) in a video stream.\n"
+                  "You can use Haar or LBP features.\n\n" );
+    parser.printMessage();
+
+    String face_cascade_name = samples::findFile( parser.get<String>("face_cascade") );
+    String eyes_cascade_name = samples::findFile( parser.get<String>("eyes_cascade") );
+
+    //-- 1. Load the cascades
+    if( !face_cascade.load( face_cascade_name ) )
+    {
+        cout << "--(!)Error loading face cascade\n";
+        return -1;
+    };
+    if( !eyes_cascade.load( eyes_cascade_name ) )
+    {
+        cout << "--(!)Error loading eyes cascade\n";
+        return -1;
+    };
+
+    int camera_device = parser.get<int>("camera");
+    VideoCapture capture;
+    //-- 2. Read the video stream
+    capture.open( camera_device );
+    if ( ! capture.isOpened() )
+    {
+        cout << "--(!)Error opening video capture\n";
+        return -1;
+    }
+
+    Mat frame;
+    while ( capture.read(frame) )
+    {
+        if( frame.empty() )
+        {
+            cout << "--(!) No captured frame -- Break!\n";
+            break;
+        }
+
+        //-- 3. Apply the classifier to the frame
+        detectAndDisplay( frame );
+
+        if( waitKey(10) == 27 )
+        {
+            break; // escape
+        }
+    }
+    return 0;
+}
+
+/** @function detectAndDisplay */
+void detectAndDisplay( Mat frame )
+{
+    Mat frame_gray;
+    cvtColor( frame, frame_gray, COLOR_BGR2GRAY );
+    equalizeHist( frame_gray, frame_gray );
+
+    //-- Detect faces
+    std::vector<Rect> faces;
+    face_cascade.detectMultiScale( frame_gray, faces );
+
+    for ( size_t i = 0; i < faces.size(); i++ )
+    {
+        Point center( faces[i].x + faces[i].width/2, faces[i].y + faces[i].height/2 );
+        ellipse( frame, center, Size( faces[i].width/2, faces[i].height/2 ), 0, 0, 360, Scalar( 255, 0, 255 ), 4 );
+
+        Mat faceROI = frame_gray( faces[i] );
+
+        //-- In each face, detect eyes
+        std::vector<Rect> eyes;
+        eyes_cascade.detectMultiScale( faceROI, eyes );
+
+        for ( size_t j = 0; j < eyes.size(); j++ )
+        {
+            Point eye_center( faces[i].x + eyes[j].x + eyes[j].width/2, faces[i].y + eyes[j].y + eyes[j].height/2 );
+            int radius = cvRound( (eyes[j].width + eyes[j].height)*0.25 );
+            circle( frame, eye_center, radius, Scalar( 255, 0, 0 ), 4 );
+        }
+    }
+
+    //-- Show what you got
+    imshow( "Capture - Face detection", frame );
+}

modules/xobjdetect/samples/peopledetect.cpp

@@ -0,0 +1,129 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html
+
+#include <opencv2/xobjdetect.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/videoio.hpp>
+#include <iostream>
+#include <iomanip>
+
+using namespace cv;
+using namespace std;
+
+class Detector
+{
+    enum Mode { Default, Daimler } m;
+    HOGDescriptor hog, hog_d;
+public:
+    Detector() : m(Default), hog(), hog_d(Size(48, 96), Size(16, 16), Size(8, 8), Size(8, 8), 9)
+    {
+        hog.setSVMDetector(HOGDescriptor::getDefaultPeopleDetector());
+        hog_d.setSVMDetector(HOGDescriptor::getDaimlerPeopleDetector());
+    }
+    void toggleMode() { m = (m == Default ? Daimler : Default); }
+    string modeName() const { return (m == Default ? "Default" : "Daimler"); }
+    vector<Rect> detect(InputArray img)
+    {
+        // Run the detector with default parameters. to get a higher hit-rate
+        // (and more false alarms, respectively), decrease the hitThreshold and
+        // groupThreshold (set groupThreshold to 0 to turn off the grouping completely).
+        vector<Rect> found;
+        if (m == Default)
+            hog.detectMultiScale(img, found, 0, Size(8,8), Size(), 1.05, 2, false);
+        else if (m == Daimler)
+            hog_d.detectMultiScale(img, found, 0, Size(8,8), Size(), 1.05, 2, true);
+        return found;
+    }
+    void adjustRect(Rect & r) const
+    {
+        // The HOG detector returns slightly larger rectangles than the real objects,
+        // so we slightly shrink the rectangles to get a nicer output.
+        r.x += cvRound(r.width*0.1);
+        r.width = cvRound(r.width*0.8);
+        r.y += cvRound(r.height*0.07);
+        r.height = cvRound(r.height*0.8);
+    }
+};
+
+static const string keys = "{ help h   |   | print help message }"
+                           "{ camera c | 0 | capture video from camera (device index starting from 0) }"
+                           "{ video v  |   | use video as input }";
+
+int main(int argc, char** argv)
+{
+    CommandLineParser parser(argc, argv, keys);
+    parser.about("This sample demonstrates the use of the HoG descriptor.");
+    if (parser.has("help"))
+    {
+        parser.printMessage();
+        return 0;
+    }
+    int camera = parser.get<int>("camera");
+    string file = parser.get<string>("video");
+    if (!parser.check())
+    {
+        parser.printErrors();
+        return 1;
+    }
+
+    VideoCapture cap;
+    if (file.empty())
+        cap.open(camera);
+    else
+    {
+        file = samples::findFileOrKeep(file);
+        cap.open(file);
+    }
+    if (!cap.isOpened())
+    {
+        cout << "Can not open video stream: '" << (file.empty() ? "<camera>" : file) << "'" << endl;
+        return 2;
+    }
+
+    cout << "Press 'q' or <ESC> to quit." << endl;
+    cout << "Press <space> to toggle between Default and Daimler detector" << endl;
+    Detector detector;
+    Mat frame;
+    for (;;)
+    {
+        cap >> frame;
+        if (frame.empty())
+        {
+            cout << "Finished reading: empty frame" << endl;
+            break;
+        }
+        int64 t = getTickCount();
+        vector<Rect> found = detector.detect(frame);
+        t = getTickCount() - t;
+
+        // show the window
+        {
+            ostringstream buf;
+            buf << "Mode: " << detector.modeName() << " ||| "
+                << "FPS: " << fixed << setprecision(1) << (getTickFrequency() / (double)t);
+            putText(frame, buf.str(), Point(10, 30), FONT_HERSHEY_PLAIN, 2.0, Scalar(0, 0, 255), 2, LINE_AA);
+        }
+        for (vector<Rect>::iterator i = found.begin(); i != found.end(); ++i)
+        {
+            Rect &r = *i;
+            detector.adjustRect(r);
+            rectangle(frame, r.tl(), r.br(), cv::Scalar(0, 255, 0), 2);
+        }
+        imshow("People detector", frame);
+
+        // interact with user
+        const char key = (char)waitKey(1);
+        if (key == 27 || key == 'q') // ESC
+        {
+            cout << "Exit requested" << endl;
+            break;
+        }
+        else if (key == ' ')
+        {
+            detector.toggleMode();
+        }
+    }
+    return 0;
+}

modules/xobjdetect/samples/python/common.py

@@ -0,0 +1,224 @@
+#!/usr/bin/env python
+
+'''
+This module contains some common routines used by other samples.
+'''
+
+from functools import reduce
+import numpy as np
+import cv2 as cv
+
+# built-in modules
+import os
+import itertools as it
+from contextlib import contextmanager
+
+image_extensions = ['.bmp', '.jpg', '.jpeg', '.png', '.tif', '.tiff', '.pbm', '.pgm', '.ppm']
+
+class Bunch(object):
+    def __init__(self, **kw):
+        self.__dict__.update(kw)
+    def __str__(self):
+        return str(self.__dict__)
+
+def splitfn(fn):
+    path, fn = os.path.split(fn)
+    name, ext = os.path.splitext(fn)
+    return path, name, ext
+
+def anorm2(a):
+    return (a*a).sum(-1)
+def anorm(a):
+    return np.sqrt( anorm2(a) )
+
+def homotrans(H, x, y):
+    xs = H[0, 0]*x + H[0, 1]*y + H[0, 2]
+    ys = H[1, 0]*x + H[1, 1]*y + H[1, 2]
+    s  = H[2, 0]*x + H[2, 1]*y + H[2, 2]
+    return xs/s, ys/s
+
+def to_rect(a):
+    a = np.ravel(a)
+    if len(a) == 2:
+        a = (0, 0, a[0], a[1])
+    return np.array(a, np.float64).reshape(2, 2)
+
+def rect2rect_mtx(src, dst):
+    src, dst = to_rect(src), to_rect(dst)
+    cx, cy = (dst[1] - dst[0]) / (src[1] - src[0])
+    tx, ty = dst[0] - src[0] * (cx, cy)
+    M = np.float64([[ cx,  0, tx],
+                    [  0, cy, ty],
+                    [  0,  0,  1]])
+    return M
+
+
+def lookat(eye, target, up = (0, 0, 1)):
+    fwd = np.asarray(target, np.float64) - eye
+    fwd /= anorm(fwd)
+    right = np.cross(fwd, up)
+    right /= anorm(right)
+    down = np.cross(fwd, right)
+    R = np.float64([right, down, fwd])
+    tvec = -np.dot(R, eye)
+    return R, tvec
+
+def mtx2rvec(R):
+    w, u, vt = cv.SVDecomp(R - np.eye(3))
+    p = vt[0] + u[:,0]*w[0]    # same as np.dot(R, vt[0])
+    c = np.dot(vt[0], p)
+    s = np.dot(vt[1], p)
+    axis = np.cross(vt[0], vt[1])
+    return axis * np.arctan2(s, c)
+
+def draw_str(dst, target, s):
+    x, y = target
+    cv.putText(dst, s, (x+1, y+1), cv.FONT_HERSHEY_PLAIN, 1.0, (0, 0, 0), thickness = 2, lineType=cv.LINE_AA)
+    cv.putText(dst, s, (x, y), cv.FONT_HERSHEY_PLAIN, 1.0, (255, 255, 255), lineType=cv.LINE_AA)
+
+class Sketcher:
+    def __init__(self, windowname, dests, colors_func):
+        self.prev_pt = None
+        self.windowname = windowname
+        self.dests = dests
+        self.colors_func = colors_func
+        self.dirty = False
+        self.show()
+        cv.setMouseCallback(self.windowname, self.on_mouse)
+
+    def show(self):
+        cv.imshow(self.windowname, self.dests[0])
+
+    def on_mouse(self, event, x, y, flags, param):
+        pt = (x, y)
+        if event == cv.EVENT_LBUTTONDOWN:
+            self.prev_pt = pt
+        elif event == cv.EVENT_LBUTTONUP:
+            self.prev_pt = None
+
+        if self.prev_pt and flags & cv.EVENT_FLAG_LBUTTON:
+            for dst, color in zip(self.dests, self.colors_func()):
+                cv.line(dst, self.prev_pt, pt, color, 5)
+            self.dirty = True
+            self.prev_pt = pt
+            self.show()
+
+
+# palette data from matplotlib/_cm.py
+_jet_data =   {'red':   ((0., 0, 0), (0.35, 0, 0), (0.66, 1, 1), (0.89,1, 1),
+                         (1, 0.5, 0.5)),
+               'green': ((0., 0, 0), (0.125,0, 0), (0.375,1, 1), (0.64,1, 1),
+                         (0.91,0,0), (1, 0, 0)),
+               'blue':  ((0., 0.5, 0.5), (0.11, 1, 1), (0.34, 1, 1), (0.65,0, 0),
+                         (1, 0, 0))}
+
+cmap_data = { 'jet' : _jet_data }
+
+def make_cmap(name, n=256):
+    data = cmap_data[name]
+    xs = np.linspace(0.0, 1.0, n)
+    channels = []
+    eps = 1e-6
+    for ch_name in ['blue', 'green', 'red']:
+        ch_data = data[ch_name]
+        xp, yp = [], []
+        for x, y1, y2 in ch_data:
+            xp += [x, x+eps]
+            yp += [y1, y2]
+        ch = np.interp(xs, xp, yp)
+        channels.append(ch)
+    return np.uint8(np.array(channels).T*255)
+
+def nothing(*arg, **kw):
+    pass
+
+def clock():
+    return cv.getTickCount() / cv.getTickFrequency()
+
+@contextmanager
+def Timer(msg):
+    print(msg, '...',)
+    start = clock()
+    try:
+        yield
+    finally:
+        print("%.2f ms" % ((clock()-start)*1000))
+
+class StatValue:
+    def __init__(self, smooth_coef = 0.5):
+        self.value = None
+        self.smooth_coef = smooth_coef
+    def update(self, v):
+        if self.value is None:
+            self.value = v
+        else:
+            c = self.smooth_coef
+            self.value = c * self.value + (1.0-c) * v
+
+class RectSelector:
+    def __init__(self, win, callback):
+        self.win = win
+        self.callback = callback
+        cv.setMouseCallback(win, self.onmouse)
+        self.drag_start = None
+        self.drag_rect = None
+    def onmouse(self, event, x, y, flags, param):
+        x, y = np.int16([x, y]) # BUG
+        if event == cv.EVENT_LBUTTONDOWN:
+            self.drag_start = (x, y)
+            return
+        if self.drag_start:
+            if flags & cv.EVENT_FLAG_LBUTTON:
+                xo, yo = self.drag_start
+                x0, y0 = np.minimum([xo, yo], [x, y])
+                x1, y1 = np.maximum([xo, yo], [x, y])
+                self.drag_rect = None
+                if x1-x0 > 0 and y1-y0 > 0:
+                    self.drag_rect = (x0, y0, x1, y1)
+            else:
+                rect = self.drag_rect
+                self.drag_start = None
+                self.drag_rect = None
+                if rect:
+                    self.callback(rect)
+    def draw(self, vis):
+        if not self.drag_rect:
+            return False
+        x0, y0, x1, y1 = self.drag_rect
+        cv.rectangle(vis, (x0, y0), (x1, y1), (0, 255, 0), 2)
+        return True
+    @property
+    def dragging(self):
+        return self.drag_rect is not None
+
+
+def grouper(n, iterable, fillvalue=None):
+    '''grouper(3, 'ABCDEFG', 'x') --> ABC DEF Gxx'''
+    args = [iter(iterable)] * n
+    output = it.zip_longest(fillvalue=fillvalue, *args)
+    return output
+
+def mosaic(w, imgs):
+    '''Make a grid from images.
+
+    w    -- number of grid columns
+    imgs -- images (must have same size and format)
+    '''
+    imgs = iter(imgs)
+    img0 = next(imgs)
+    pad = np.zeros_like(img0)
+    imgs = it.chain([img0], imgs)
+    rows = grouper(w, imgs, pad)
+    return np.vstack(list(map(np.hstack, rows)))
+
+def getsize(img):
+    h, w = img.shape[:2]
+    return w, h
+
+def mdot(*args):
+    return reduce(np.dot, args)
+
+def draw_keypoints(vis, keypoints, color = (0, 255, 255)):
+    for kp in keypoints:
+        x, y = kp.pt
+        cv.circle(vis, (int(x), int(y)), 2, color)

modules/xobjdetect/samples/python/facedetect.py

@@ -0,0 +1,79 @@
+#!/usr/bin/env python
+
+'''
+face detection using haar cascades
+
+USAGE:
+    facedetect.py [--cascade <cascade_fn>] [--nested-cascade <cascade_fn>] [<video_source>]
+'''
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+
+# local modules
+from video import create_capture
+from common import clock, draw_str
+
+
+def detect(img, cascade):
+    rects = cascade.detectMultiScale(img, scaleFactor=1.3, minNeighbors=4, minSize=(30, 30),
+                                     flags=cv.CASCADE_SCALE_IMAGE)
+    if len(rects) == 0:
+        return []
+    rects[:,2:] += rects[:,:2]
+    return rects
+
+def draw_rects(img, rects, color):
+    for x1, y1, x2, y2 in rects:
+        cv.rectangle(img, (x1, y1), (x2, y2), color, 2)
+
+def main():
+    import sys, getopt
+
+    args, video_src = getopt.getopt(sys.argv[1:], '', ['cascade=', 'nested-cascade='])
+    try:
+        video_src = video_src[0]
+    except:
+        video_src = 0
+    args = dict(args)
+    cascade_fn = args.get('--cascade', "haarcascades/haarcascade_frontalface_alt.xml")
+    nested_fn  = args.get('--nested-cascade', "haarcascades/haarcascade_eye.xml")
+
+    cascade = cv.CascadeClassifier(cv.samples.findFile(cascade_fn))
+    nested = cv.CascadeClassifier(cv.samples.findFile(nested_fn))
+
+    cam = create_capture(video_src, fallback='synth:bg={}:noise=0.05'.format(cv.samples.findFile('lena.jpg')))
+
+    while True:
+        _ret, img = cam.read()
+        gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
+        gray = cv.equalizeHist(gray)
+
+        t = clock()
+        rects = detect(gray, cascade)
+        vis = img.copy()
+        draw_rects(vis, rects, (0, 255, 0))
+        if not nested.empty():
+            for x1, y1, x2, y2 in rects:
+                roi = gray[y1:y2, x1:x2]
+                vis_roi = vis[y1:y2, x1:x2]
+                subrects = detect(roi.copy(), nested)
+                draw_rects(vis_roi, subrects, (255, 0, 0))
+        dt = clock() - t
+
+        draw_str(vis, (20, 20), 'time: %.1f ms' % (dt*1000))
+        cv.imshow('facedetect', vis)
+
+        if cv.waitKey(5) == 27:
+            break
+
+    print('Done')
+
+
+if __name__ == '__main__':
+    print(__doc__)
+    main()
+    cv.destroyAllWindows()

modules/xobjdetect/samples/python/peopledetect.py

@@ -0,0 +1,76 @@
+#!/usr/bin/env python
+
+'''
+example to detect upright people in images using HOG features
+
+Usage:
+    peopledetect.py <image_names>
+
+Press any key to continue, ESC to stop.
+'''
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+
+
+def inside(r, q):
+    rx, ry, rw, rh = r
+    qx, qy, qw, qh = q
+    return rx > qx and ry > qy and rx + rw < qx + qw and ry + rh < qy + qh
+
+
+def draw_detections(img, rects, thickness = 1):
+    for x, y, w, h in rects:
+        # the HOG detector returns slightly larger rectangles than the real objects.
+        # so we slightly shrink the rectangles to get a nicer output.
+        pad_w, pad_h = int(0.15*w), int(0.05*h)
+        cv.rectangle(img, (x+pad_w, y+pad_h), (x+w-pad_w, y+h-pad_h), (0, 255, 0), thickness)
+
+
+def main():
+    import sys
+    from glob import glob
+    import itertools as it
+
+    hog = cv.HOGDescriptor()
+    hog.setSVMDetector( cv.HOGDescriptor_getDefaultPeopleDetector() )
+
+    default = [cv.samples.findFile('basketball2.png')] if len(sys.argv[1:]) == 0 else []
+
+    for fn in it.chain(*map(glob, default + sys.argv[1:])):
+        print(fn, ' - ',)
+        try:
+            img = cv.imread(fn)
+            if img is None:
+                print('Failed to load image file:', fn)
+                continue
+        except:
+            print('loading error')
+            continue
+
+        found, _w = hog.detectMultiScale(img, winStride=(8,8), padding=(32,32), scale=1.05)
+        found_filtered = []
+        for ri, r in enumerate(found):
+            for qi, q in enumerate(found):
+                if ri != qi and inside(r, q):
+                    break
+            else:
+                found_filtered.append(r)
+        draw_detections(img, found)
+        draw_detections(img, found_filtered, 3)
+        print('%d (%d) found' % (len(found_filtered), len(found)))
+        cv.imshow('img', img)
+        ch = cv.waitKey()
+        if ch == 27:
+            break
+
+    print('Done')
+
+
+if __name__ == '__main__':
+    print(__doc__)
+    main()
+    cv.destroyAllWindows()

modules/xobjdetect/samples/python/tst_scene_render.py

@@ -0,0 +1,121 @@
+#!/usr/bin/env python
+
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+
+from numpy import pi, sin, cos
+
+
+defaultSize = 512
+
+class TestSceneRender():
+
+    def __init__(self, bgImg = None, fgImg = None,
+        deformation = False, speed = 0.25, **params):
+        self.time = 0.0
+        self.timeStep = 1.0 / 30.0
+        self.foreground = fgImg
+        self.deformation = deformation
+        self.speed = speed
+
+        if bgImg is not None:
+            self.sceneBg = bgImg.copy()
+        else:
+            self.sceneBg = np.zeros((defaultSize, defaultSize,3), np.uint8)
+
+        self.w = self.sceneBg.shape[0]
+        self.h = self.sceneBg.shape[1]
+
+        if fgImg is not None:
+            self.foreground = fgImg.copy()
+            self.center = self.currentCenter = (int(self.w/2 - fgImg.shape[0]/2), int(self.h/2 - fgImg.shape[1]/2))
+
+            self.xAmpl = self.sceneBg.shape[0] - (self.center[0] + fgImg.shape[0])
+            self.yAmpl = self.sceneBg.shape[1] - (self.center[1] + fgImg.shape[1])
+
+        self.initialRect = np.array([ (self.h/2, self.w/2), (self.h/2, self.w/2 + self.w/10),
+         (self.h/2 + self.h/10, self.w/2 + self.w/10), (self.h/2 + self.h/10, self.w/2)]).astype(int)
+        self.currentRect = self.initialRect
+
+    def getXOffset(self, time):
+        return int( self.xAmpl*cos(time*self.speed))
+
+
+    def getYOffset(self, time):
+        return int(self.yAmpl*sin(time*self.speed))
+
+    def setInitialRect(self, rect):
+        self.initialRect = rect
+
+    def getRectInTime(self, time):
+
+        if self.foreground is not None:
+            tmp = np.array(self.center) + np.array((self.getXOffset(time), self.getYOffset(time)))
+            x0, y0 = tmp
+            x1, y1 = tmp + self.foreground.shape[0:2]
+            return np.array([y0, x0, y1, x1])
+        else:
+            x0, y0 = self.initialRect[0] + np.array((self.getXOffset(time), self.getYOffset(time)))
+            x1, y1 = self.initialRect[2] + np.array((self.getXOffset(time), self.getYOffset(time)))
+            return np.array([y0, x0, y1, x1])
+
+    def getCurrentRect(self):
+
+        if self.foreground is not None:
+
+            x0 = self.currentCenter[0]
+            y0 = self.currentCenter[1]
+            x1 = self.currentCenter[0] + self.foreground.shape[0]
+            y1 = self.currentCenter[1] + self.foreground.shape[1]
+            return np.array([y0, x0, y1, x1])
+        else:
+            x0, y0 = self.currentRect[0]
+            x1, y1 = self.currentRect[2]
+            return np.array([x0, y0, x1, y1])
+
+    def getNextFrame(self):
+        img = self.sceneBg.copy()
+
+        if self.foreground is not None:
+            self.currentCenter = (self.center[0] + self.getXOffset(self.time), self.center[1] + self.getYOffset(self.time))
+            img[self.currentCenter[0]:self.currentCenter[0]+self.foreground.shape[0],
+             self.currentCenter[1]:self.currentCenter[1]+self.foreground.shape[1]] = self.foreground
+        else:
+            self.currentRect = self.initialRect + int( 30*cos(self.time*self.speed) + 50*sin(self.time*self.speed))
+            if self.deformation:
+                self.currentRect[1:3] += int(self.h/20*cos(self.time))
+            cv.fillConvexPoly(img, self.currentRect, (0, 0, 255))
+
+        self.time += self.timeStep
+        return img
+
+    def resetTime(self):
+        self.time = 0.0
+
+
+def main():
+    backGr = cv.imread(cv.samples.findFile('graf1.png'))
+    fgr = cv.imread(cv.samples.findFile('box.png'))
+
+    render = TestSceneRender(backGr, fgr)
+
+    while True:
+
+        img = render.getNextFrame()
+        cv.imshow('img', img)
+
+        ch = cv.waitKey(3)
+        if  ch == 27:
+            break
+
+    print('Done')
+
+
+if __name__ == '__main__':
+    print(__doc__)
+    main()
+    cv.destroyAllWindows()

modules/xobjdetect/samples/python/tutorial_code/objectDetection/cascade_classifier/objectDetection.py

@@ -0,0 +1,61 @@
+from __future__ import print_function
+import cv2 as cv
+import argparse
+
+def detectAndDisplay(frame):
+    frame_gray = cv.cvtColor(frame, cv.COLOR_BGR2GRAY)
+    frame_gray = cv.equalizeHist(frame_gray)
+
+    #-- Detect faces
+    faces = face_cascade.detectMultiScale(frame_gray)
+    for (x,y,w,h) in faces:
+        center = (x + w//2, y + h//2)
+        frame = cv.ellipse(frame, center, (w//2, h//2), 0, 0, 360, (255, 0, 255), 4)
+
+        faceROI = frame_gray[y:y+h,x:x+w]
+        #-- In each face, detect eyes
+        eyes = eyes_cascade.detectMultiScale(faceROI)
+        for (x2,y2,w2,h2) in eyes:
+            eye_center = (x + x2 + w2//2, y + y2 + h2//2)
+            radius = int(round((w2 + h2)*0.25))
+            frame = cv.circle(frame, eye_center, radius, (255, 0, 0 ), 4)
+
+    cv.imshow('Capture - Face detection', frame)
+
+parser = argparse.ArgumentParser(description='Code for Cascade Classifier tutorial.')
+parser.add_argument('--face_cascade', help='Path to face cascade.', default='data/haarcascades/haarcascade_frontalface_alt.xml')
+parser.add_argument('--eyes_cascade', help='Path to eyes cascade.', default='data/haarcascades/haarcascade_eye_tree_eyeglasses.xml')
+parser.add_argument('--camera', help='Camera divide number.', type=int, default=0)
+args = parser.parse_args()
+
+face_cascade_name = args.face_cascade
+eyes_cascade_name = args.eyes_cascade
+
+face_cascade = cv.CascadeClassifier()
+eyes_cascade = cv.CascadeClassifier()
+
+#-- 1. Load the cascades
+if not face_cascade.load(cv.samples.findFile(face_cascade_name)):
+    print('--(!)Error loading face cascade')
+    exit(0)
+if not eyes_cascade.load(cv.samples.findFile(eyes_cascade_name)):
+    print('--(!)Error loading eyes cascade')
+    exit(0)
+
+camera_device = args.camera
+#-- 2. Read the video stream
+cap = cv.VideoCapture(camera_device)
+if not cap.isOpened:
+    print('--(!)Error opening video capture')
+    exit(0)
+
+while True:
+    ret, frame = cap.read()
+    if frame is None:
+        print('--(!) No captured frame -- Break!')
+        break
+
+    detectAndDisplay(frame)
+
+    if cv.waitKey(10) == 27:
+        break

modules/xobjdetect/samples/python/video.py

@@ -0,0 +1,228 @@
+#!/usr/bin/env python
+
+'''
+Video capture sample.
+
+Sample shows how VideoCapture class can be used to acquire video
+frames from a camera of a movie file. Also the sample provides
+an example of procedural video generation by an object, mimicking
+the VideoCapture interface (see Chess class).
+
+'create_capture' is a convenience function for capture creation,
+falling back to procedural video in case of error.
+
+Usage:
+    video.py [--shotdir <shot path>] [source0] [source1] ...'
+
+    sourceN is an
+     - integer number for camera capture
+     - name of video file
+     - synth:<params> for procedural video
+
+Synth examples:
+    synth:bg=lena.jpg:noise=0.1
+    synth:class=chess:bg=lena.jpg:noise=0.1:size=640x480
+
+Keys:
+    ESC    - exit
+    SPACE  - save current frame to <shot path> directory
+
+'''
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+
+import re
+
+from numpy import pi, sin, cos
+
+# local modules
+from tst_scene_render import TestSceneRender
+import common
+
+class VideoSynthBase(object):
+    def __init__(self, size=None, noise=0.0, bg = None, **params):
+        self.bg = None
+        self.frame_size = (640, 480)
+        if bg is not None:
+            self.bg = cv.imread(cv.samples.findFile(bg))
+            h, w = self.bg.shape[:2]
+            self.frame_size = (w, h)
+
+        if size is not None:
+            w, h = map(int, size.split('x'))
+            self.frame_size = (w, h)
+            self.bg = cv.resize(self.bg, self.frame_size)
+
+        self.noise = float(noise)
+
+    def render(self, dst):
+        pass
+
+    def read(self, dst=None):
+        w, h = self.frame_size
+
+        if self.bg is None:
+            buf = np.zeros((h, w, 3), np.uint8)
+        else:
+            buf = self.bg.copy()
+
+        self.render(buf)
+
+        if self.noise > 0.0:
+            noise = np.zeros((h, w, 3), np.int8)
+            cv.randn(noise, np.zeros(3), np.ones(3)*255*self.noise)
+            buf = cv.add(buf, noise, dtype=cv.CV_8UC3)
+        return True, buf
+
+    def isOpened(self):
+        return True
+
+class Book(VideoSynthBase):
+    def __init__(self, **kw):
+        super(Book, self).__init__(**kw)
+        backGr = cv.imread(cv.samples.findFile('graf1.png'))
+        fgr = cv.imread(cv.samples.findFile('box.png'))
+        self.render = TestSceneRender(backGr, fgr, speed = 1)
+
+    def read(self, dst=None):
+        noise = np.zeros(self.render.sceneBg.shape, np.int8)
+        cv.randn(noise, np.zeros(3), np.ones(3)*255*self.noise)
+
+        return True, cv.add(self.render.getNextFrame(), noise, dtype=cv.CV_8UC3)
+
+class Cube(VideoSynthBase):
+    def __init__(self, **kw):
+        super(Cube, self).__init__(**kw)
+        self.render = TestSceneRender(cv.imread(cv.samples.findFile('pca_test1.jpg')), deformation = True,  speed = 1)
+
+    def read(self, dst=None):
+        noise = np.zeros(self.render.sceneBg.shape, np.int8)
+        cv.randn(noise, np.zeros(3), np.ones(3)*255*self.noise)
+
+        return True, cv.add(self.render.getNextFrame(), noise, dtype=cv.CV_8UC3)
+
+class Chess(VideoSynthBase):
+    def __init__(self, **kw):
+        super(Chess, self).__init__(**kw)
+
+        w, h = self.frame_size
+
+        self.grid_size = sx, sy = 10, 7
+        white_quads = []
+        black_quads = []
+        for i, j in np.ndindex(sy, sx):
+            q = [[j, i, 0], [j+1, i, 0], [j+1, i+1, 0], [j, i+1, 0]]
+            [white_quads, black_quads][(i + j) % 2].append(q)
+        self.white_quads = np.float32(white_quads)
+        self.black_quads = np.float32(black_quads)
+
+        fx = 0.9
+        self.K = np.float64([[fx*w, 0, 0.5*(w-1)],
+                        [0, fx*w, 0.5*(h-1)],
+                        [0.0,0.0,      1.0]])
+
+        self.dist_coef = np.float64([-0.2, 0.1, 0, 0])
+        self.t = 0
+
+    def draw_quads(self, img, quads, color = (0, 255, 0)):
+        img_quads = cv.projectPoints(quads.reshape(-1, 3), self.rvec, self.tvec, self.K, self.dist_coef) [0]
+        img_quads.shape = quads.shape[:2] + (2,)
+        for q in img_quads:
+            cv.fillConvexPoly(img, np.int32(q*4), color, cv.LINE_AA, shift=2)
+
+    def render(self, dst):
+        t = self.t
+        self.t += 1.0/30.0
+
+        sx, sy = self.grid_size
+        center = np.array([0.5*sx, 0.5*sy, 0.0])
+        phi = pi/3 + sin(t*3)*pi/8
+        c, s = cos(phi), sin(phi)
+        ofs = np.array([sin(1.2*t), cos(1.8*t), 0]) * sx * 0.2
+        eye_pos = center + np.array([cos(t)*c, sin(t)*c, s]) * 15.0 + ofs
+        target_pos = center + ofs
+
+        R, self.tvec = common.lookat(eye_pos, target_pos)
+        self.rvec = common.mtx2rvec(R)
+
+        self.draw_quads(dst, self.white_quads, (245, 245, 245))
+        self.draw_quads(dst, self.black_quads, (10, 10, 10))
+
+
+classes = dict(chess=Chess, book=Book, cube=Cube)
+
+presets = dict(
+    empty = 'synth:',
+    lena = 'synth:bg=lena.jpg:noise=0.1',
+    chess = 'synth:class=chess:bg=lena.jpg:noise=0.1:size=640x480',
+    book = 'synth:class=book:bg=graf1.png:noise=0.1:size=640x480',
+    cube = 'synth:class=cube:bg=pca_test1.jpg:noise=0.0:size=640x480'
+)
+
+
+def create_capture(source = 0, fallback = presets['chess']):
+    '''source: <int> or '<int>|<filename>|synth [:<param_name>=<value> [:...]]'
+    '''
+    source = str(source).strip()
+
+    # Win32: handle drive letter ('c:', ...)
+    source = re.sub(r'(^|=)([a-zA-Z]):([/\\a-zA-Z0-9])', r'\1?disk\2?\3', source)
+    chunks = source.split(':')
+    chunks = [re.sub(r'\?disk([a-zA-Z])\?', r'\1:', s) for s in chunks]
+
+    source = chunks[0]
+    try: source = int(source)
+    except ValueError: pass
+    params = dict( s.split('=') for s in chunks[1:] )
+
+    cap = None
+    if source == 'synth':
+        Class = classes.get(params.get('class', None), VideoSynthBase)
+        try: cap = Class(**params)
+        except: pass
+    else:
+        cap = cv.VideoCapture(source)
+        if 'size' in params:
+            w, h = map(int, params['size'].split('x'))
+            cap.set(cv.CAP_PROP_FRAME_WIDTH, w)
+            cap.set(cv.CAP_PROP_FRAME_HEIGHT, h)
+    if cap is None or not cap.isOpened():
+        print('Warning: unable to open video source: ', source)
+        if fallback is not None:
+            return create_capture(fallback, None)
+    return cap
+
+if __name__ == '__main__':
+    import sys
+    import getopt
+
+    print(__doc__)
+
+    args, sources = getopt.getopt(sys.argv[1:], '', 'shotdir=')
+    args = dict(args)
+    shotdir = args.get('--shotdir', '.')
+    if len(sources) == 0:
+        sources = [ 0 ]
+
+    caps = list(map(create_capture, sources))
+    shot_idx = 0
+    while True:
+        imgs = []
+        for i, cap in enumerate(caps):
+            ret, img = cap.read()
+            imgs.append(img)
+            cv.imshow('capture %d' % i, img)
+        ch = cv.waitKey(1)
+        if ch == 27:
+            break
+        if ch == ord(' '):
+            for i, img in enumerate(imgs):
+                fn = '%s/shot_%d_%03d.bmp' % (shotdir, i, shot_idx)
+                cv.imwrite(fn, img)
+                print(fn, 'saved')
+            shot_idx += 1
+    cv.destroyAllWindows()

modules/xobjdetect/samples/smiledetect.cpp

@@ -0,0 +1,215 @@
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/videoio.hpp"
+#include <iostream>
+
+using namespace std;
+using namespace cv;
+
+static void help(const char** argv)
+{
+    cout << "\nThis program demonstrates the smile detector.\n"
+            "Usage:\n" <<
+            argv[0] << " [--cascade=<cascade_path> this is the frontal face classifier]\n"
+            "   [--smile-cascade=[<smile_cascade_path>]]\n"
+            "   [--scale=<image scale greater or equal to 1, try 2.0 for example. The larger the faster the processing>]\n"
+            "   [--try-flip]\n"
+            "   [video_filename|camera_index]\n\n"
+            "Example:\n" <<
+            argv[0] << " --cascade=\"data/haarcascades/haarcascade_frontalface_alt.xml\" --smile-cascade=\"data/haarcascades/haarcascade_smile.xml\" --scale=2.0\n\n"
+            "During execution:\n\tHit any key to quit.\n"
+            "\tUsing OpenCV version " << CV_VERSION << "\n" << endl;
+}
+
+void detectAndDraw( Mat& img, CascadeClassifier& cascade,
+                    CascadeClassifier& nestedCascade,
+                    double scale, bool tryflip );
+
+string cascadeName;
+string nestedCascadeName;
+
+int main( int argc, const char** argv )
+{
+    VideoCapture capture;
+    Mat frame, image;
+    string inputName;
+    bool tryflip;
+
+    help(argv);
+
+    CascadeClassifier cascade, nestedCascade;
+    double scale;
+    cv::CommandLineParser parser(argc, argv,
+        "{help h||}{scale|1|}"
+        "{cascade|data/haarcascades/haarcascade_frontalface_alt.xml|}"
+        "{smile-cascade|data/haarcascades/haarcascade_smile.xml|}"
+        "{try-flip||}{@input||}");
+    if (parser.has("help"))
+    {
+        help(argv);
+        return 0;
+    }
+    cascadeName = samples::findFile(parser.get<string>("cascade"));
+    nestedCascadeName = samples::findFile(parser.get<string>("smile-cascade"));
+    tryflip = parser.has("try-flip");
+    inputName = parser.get<string>("@input");
+    scale = parser.get<int>("scale");
+    if (!parser.check())
+    {
+        help(argv);
+        return 1;
+    }
+    if (scale < 1)
+        scale = 1;
+    if( !cascade.load( cascadeName ) )
+    {
+        cerr << "ERROR: Could not load face cascade" << endl;
+        help(argv);
+        return -1;
+    }
+    if( !nestedCascade.load( nestedCascadeName ) )
+    {
+        cerr << "ERROR: Could not load smile cascade" << endl;
+        help(argv);
+        return -1;
+    }
+    if( inputName.empty() || (isdigit(inputName[0]) && inputName.size() == 1) )
+    {
+        int c = inputName.empty() ? 0 : inputName[0] - '0' ;
+        if(!capture.open(c))
+            cout << "Capture from camera #" <<  c << " didn't work" << endl;
+    }
+    else if( inputName.size() )
+    {
+        inputName = samples::findFileOrKeep(inputName);
+        if(!capture.open( inputName ))
+            cout << "Could not read " << inputName << endl;
+    }
+
+    if( capture.isOpened() )
+    {
+        cout << "Video capturing has been started ..." << endl;
+        cout << endl << "NOTE: Smile intensity will only be valid after a first smile has been detected" << endl;
+
+        for(;;)
+        {
+            capture >> frame;
+            if( frame.empty() )
+                break;
+
+            Mat frame1 = frame.clone();
+            detectAndDraw( frame1, cascade, nestedCascade, scale, tryflip );
+
+            char c = (char)waitKey(10);
+            if( c == 27 || c == 'q' || c == 'Q' )
+                break;
+        }
+    }
+    else
+    {
+        cerr << "ERROR: Could not initiate capture" << endl;
+        help(argv);
+        return -1;
+    }
+
+    return 0;
+}
+
+void detectAndDraw( Mat& img, CascadeClassifier& cascade,
+                    CascadeClassifier& nestedCascade,
+                    double scale, bool tryflip)
+{
+    vector<Rect> faces, faces2;
+    const static Scalar colors[] =
+    {
+        Scalar(255,0,0),
+        Scalar(255,128,0),
+        Scalar(255,255,0),
+        Scalar(0,255,0),
+        Scalar(0,128,255),
+        Scalar(0,255,255),
+        Scalar(0,0,255),
+        Scalar(255,0,255)
+    };
+    Mat gray, smallImg;
+
+    cvtColor( img, gray, COLOR_BGR2GRAY );
+
+    double fx = 1 / scale;
+    resize( gray, smallImg, Size(), fx, fx, INTER_LINEAR_EXACT );
+    equalizeHist( smallImg, smallImg );
+
+    cascade.detectMultiScale( smallImg, faces,
+        1.1, 2, 0
+        //|CASCADE_FIND_BIGGEST_OBJECT
+        //|CASCADE_DO_ROUGH_SEARCH
+        |CASCADE_SCALE_IMAGE,
+        Size(30, 30) );
+    if( tryflip )
+    {
+        flip(smallImg, smallImg, 1);
+        cascade.detectMultiScale( smallImg, faces2,
+                                 1.1, 2, 0
+                                 //|CASCADE_FIND_BIGGEST_OBJECT
+                                 //|CASCADE_DO_ROUGH_SEARCH
+                                 |CASCADE_SCALE_IMAGE,
+                                 Size(30, 30) );
+        for( vector<Rect>::const_iterator r = faces2.begin(); r != faces2.end(); ++r )
+        {
+            faces.push_back(Rect(smallImg.cols - r->x - r->width, r->y, r->width, r->height));
+        }
+    }
+
+    for ( size_t i = 0; i < faces.size(); i++ )
+    {
+        Rect r = faces[i];
+        Mat smallImgROI;
+        vector<Rect> nestedObjects;
+        Point center;
+        Scalar color = colors[i%8];
+        int radius;
+
+        double aspect_ratio = (double)r.width/r.height;
+        if( 0.75 < aspect_ratio && aspect_ratio < 1.3 )
+        {
+            center.x = cvRound((r.x + r.width*0.5)*scale);
+            center.y = cvRound((r.y + r.height*0.5)*scale);
+            radius = cvRound((r.width + r.height)*0.25*scale);
+            circle( img, center, radius, color, 3, 8, 0 );
+        }
+        else
+            rectangle( img, Point(cvRound(r.x*scale), cvRound(r.y*scale)),
+                       Point(cvRound((r.x + r.width-1)*scale), cvRound((r.y + r.height-1)*scale)),
+                       color, 3, 8, 0);
+
+        const int half_height=cvRound((float)r.height/2);
+        r.y=r.y + half_height;
+        r.height = half_height-1;
+        smallImgROI = smallImg( r );
+        nestedCascade.detectMultiScale( smallImgROI, nestedObjects,
+            1.1, 0, 0
+            //|CASCADE_FIND_BIGGEST_OBJECT
+            //|CASCADE_DO_ROUGH_SEARCH
+            //|CASCADE_DO_CANNY_PRUNING
+            |CASCADE_SCALE_IMAGE,
+            Size(30, 30) );
+
+        // The number of detected neighbors depends on image size (and also illumination, etc.). The
+        // following steps use a floating minimum and maximum of neighbors. Intensity thus estimated will be
+        //accurate only after a first smile has been displayed by the user.
+        const int smile_neighbors = (int)nestedObjects.size();
+        static int max_neighbors=-1;
+        static int min_neighbors=-1;
+        if (min_neighbors == -1) min_neighbors = smile_neighbors;
+        max_neighbors = MAX(max_neighbors, smile_neighbors);
+
+        // Draw rectangle on the left side of the image reflecting smile intensity
+        float intensityZeroOne = ((float)smile_neighbors - min_neighbors) / (max_neighbors - min_neighbors + 1);
+        int rect_height = cvRound((float)img.rows * intensityZeroOne);
+        Scalar col = Scalar((float)255 * intensityZeroOne, 0, 0);
+        rectangle(img, Point(0, img.rows), Point(img.cols/10, img.rows - rect_height), col, -1);
+    }
+
+    imshow( "result", img );
+}

modules/xobjdetect/samples/ufacedetect.cpp

@@ -0,0 +1,253 @@
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/highgui.hpp"
+#include "opencv2/imgproc.hpp"
+#include "opencv2/core/ocl.hpp"
+#include <iostream>
+
+using namespace std;
+using namespace cv;
+
+static void help()
+{
+    cout << "\nThis program demonstrates the cascade recognizer. Now you can use Haar or LBP features.\n"
+            "This classifier can recognize many kinds of rigid objects, once the appropriate classifier is trained.\n"
+            "It's most known use is for faces.\n"
+            "Usage:\n"
+            "./ufacedetect [--cascade=<cascade_path> this is the primary trained classifier such as frontal face]\n"
+               "   [--nested-cascade[=nested_cascade_path this an optional secondary classifier such as eyes]]\n"
+               "   [--scale=<image scale greater or equal to 1, try 1.3 for example>]\n"
+               "   [--try-flip]\n"
+               "   [filename|camera_index]\n\n"
+            "see facedetect.cmd for one call:\n"
+            "./ufacedetect --cascade=\"../../data/haarcascades/haarcascade_frontalface_alt.xml\" --nested-cascade=\"../../data/haarcascades/haarcascade_eye_tree_eyeglasses.xml\" --scale=1.3\n\n"
+            "During execution:\n\tHit any key to quit.\n"
+            "\tUsing OpenCV version " << CV_VERSION << "\n" << endl;
+}
+
+void detectAndDraw( UMat& img, Mat& canvas, CascadeClassifier& cascade,
+                    CascadeClassifier& nestedCascade,
+                    double scale, bool tryflip );
+
+int main( int argc, const char** argv )
+{
+    VideoCapture capture;
+    UMat frame, image;
+    Mat canvas;
+
+    string inputName;
+    bool tryflip;
+
+    CascadeClassifier cascade, nestedCascade;
+    double scale;
+
+    cv::CommandLineParser parser(argc, argv,
+        "{cascade|data/haarcascades/haarcascade_frontalface_alt.xml|}"
+        "{nested-cascade|data/haarcascades/haarcascade_eye_tree_eyeglasses.xml|}"
+        "{help h ||}{scale|1|}{try-flip||}{@filename||}"
+    );
+    if (parser.has("help"))
+    {
+        help();
+        return 0;
+    }
+    string cascadeName = samples::findFile(parser.get<string>("cascade"));
+    string nestedCascadeName = samples::findFileOrKeep(parser.get<string>("nested-cascade"));
+    scale = parser.get<double>("scale");
+    tryflip = parser.has("try-flip");
+    inputName = parser.get<string>("@filename");
+    if ( !parser.check())
+    {
+        parser.printErrors();
+        help();
+        return -1;
+    }
+
+    if ( !nestedCascade.load( nestedCascadeName ) )
+        cerr << "WARNING: Could not load classifier cascade for nested objects: " << nestedCascadeName << endl;
+    if( !cascade.load( cascadeName ) )
+    {
+        cerr << "ERROR: Could not load classifier cascade: " << cascadeName << endl;
+        help();
+        return -1;
+    }
+
+    cout << "old cascade: " << (cascade.isOldFormatCascade() ? "TRUE" : "FALSE") << endl;
+
+    if( inputName.empty() || (isdigit(inputName[0]) && inputName.size() == 1) )
+    {
+        int camera = inputName.empty() ? 0 : inputName[0] - '0';
+        if(!capture.open(camera))
+            cout << "Capture from camera #" <<  camera << " didn't work" << endl;
+    }
+    else
+    {
+        inputName = samples::findFileOrKeep(inputName);
+        imread(inputName, IMREAD_COLOR).copyTo(image);
+        if( image.empty() )
+        {
+            if(!capture.open( inputName ))
+                cout << "Could not read " << inputName << endl;
+        }
+    }
+
+    if( capture.isOpened() )
+    {
+        cout << "Video capturing has been started ..." << endl;
+        for(;;)
+        {
+            capture >> frame;
+            if( frame.empty() )
+                break;
+
+            detectAndDraw( frame, canvas, cascade, nestedCascade, scale, tryflip );
+
+            char c = (char)waitKey(10);
+            if( c == 27 || c == 'q' || c == 'Q' )
+                break;
+        }
+    }
+    else
+    {
+        cout << "Detecting face(s) in " << inputName << endl;
+        if( !image.empty() )
+        {
+            detectAndDraw( image, canvas, cascade, nestedCascade, scale, tryflip );
+            waitKey(0);
+        }
+        else if( !inputName.empty() )
+        {
+            /* assume it is a text file containing the
+            list of the image filenames to be processed - one per line */
+            FILE* f = fopen( inputName.c_str(), "rt" );
+            if( f )
+            {
+                char buf[1000+1];
+                while( fgets( buf, 1000, f ) )
+                {
+                    int len = (int)strlen(buf);
+                    while( len > 0 && isspace(buf[len-1]) )
+                        len--;
+                    buf[len] = '\0';
+                    cout << "file " << buf << endl;
+                    imread(samples::findFile(buf), IMREAD_COLOR).copyTo(image);
+                    if( !image.empty() )
+                    {
+                        detectAndDraw( image, canvas, cascade, nestedCascade, scale, tryflip );
+                        char c = (char)waitKey(0);
+                        if( c == 27 || c == 'q' || c == 'Q' )
+                            break;
+                    }
+                    else
+                    {
+                        cerr << "Aw snap, couldn't read image " << buf << endl;
+                    }
+                }
+                fclose(f);
+            }
+        }
+    }
+
+    return 0;
+}
+
+void detectAndDraw( UMat& img, Mat& canvas, CascadeClassifier& cascade,
+                    CascadeClassifier& nestedCascade,
+                    double scale, bool tryflip )
+{
+    double t = 0;
+    vector<Rect> faces, faces2;
+    const static Scalar colors[] =
+    {
+        Scalar(255,0,0),
+        Scalar(255,128,0),
+        Scalar(255,255,0),
+        Scalar(0,255,0),
+        Scalar(0,128,255),
+        Scalar(0,255,255),
+        Scalar(0,0,255),
+        Scalar(255,0,255)
+    };
+    static UMat gray, smallImg;
+
+    t = (double)getTickCount();
+
+    cvtColor( img, gray, COLOR_BGR2GRAY );
+    double fx = 1 / scale;
+    resize( gray, smallImg, Size(), fx, fx, INTER_LINEAR_EXACT );
+    equalizeHist( smallImg, smallImg );
+
+    cascade.detectMultiScale( smallImg, faces,
+        1.1, 3, 0
+        //|CASCADE_FIND_BIGGEST_OBJECT
+        //|CASCADE_DO_ROUGH_SEARCH
+        |CASCADE_SCALE_IMAGE,
+        Size(30, 30) );
+    if( tryflip )
+    {
+        flip(smallImg, smallImg, 1);
+        cascade.detectMultiScale( smallImg, faces2,
+                                 1.1, 2, 0
+                                 //|CASCADE_FIND_BIGGEST_OBJECT
+                                 //|CASCADE_DO_ROUGH_SEARCH
+                                 |CASCADE_SCALE_IMAGE,
+                                 Size(30, 30) );
+        for( vector<Rect>::const_iterator r = faces2.begin(); r != faces2.end(); ++r )
+        {
+            faces.push_back(Rect(smallImg.cols - r->x - r->width, r->y, r->width, r->height));
+        }
+    }
+    t = (double)getTickCount() - t;
+    img.copyTo(canvas);
+
+    double fps = getTickFrequency()/t;
+    static double avgfps = 0;
+    static int nframes = 0;
+    nframes++;
+    double alpha = nframes > 50 ? 0.01 : 1./nframes;
+    avgfps = avgfps*(1-alpha) + fps*alpha;
+
+    putText(canvas, cv::format("OpenCL: %s, fps: %.1f", ocl::useOpenCL() ? "ON" : "OFF", avgfps), Point(50, 30),
+            FONT_HERSHEY_SIMPLEX, 0.8, Scalar(0,255,0), 2);
+
+    for ( size_t i = 0; i < faces.size(); i++ )
+    {
+        Rect r = faces[i];
+        vector<Rect> nestedObjects;
+        Point center;
+        Scalar color = colors[i%8];
+        int radius;
+
+        double aspect_ratio = (double)r.width/r.height;
+        if( 0.75 < aspect_ratio && aspect_ratio < 1.3 )
+        {
+            center.x = cvRound((r.x + r.width*0.5)*scale);
+            center.y = cvRound((r.y + r.height*0.5)*scale);
+            radius = cvRound((r.width + r.height)*0.25*scale);
+            circle( canvas, center, radius, color, 3, 8, 0 );
+        }
+        else
+            rectangle( canvas, Point(cvRound(r.x*scale), cvRound(r.y*scale)),
+                       Point(cvRound((r.x + r.width-1)*scale), cvRound((r.y + r.height-1)*scale)),
+                       color, 3, 8, 0);
+        if( nestedCascade.empty() )
+            continue;
+        UMat smallImgROI = smallImg(r);
+        nestedCascade.detectMultiScale( smallImgROI, nestedObjects,
+            1.1, 2, 0
+            //|CASCADE_FIND_BIGGEST_OBJECT
+            //|CASCADE_DO_ROUGH_SEARCH
+            //|CASCADE_DO_CANNY_PRUNING
+            |CASCADE_SCALE_IMAGE,
+            Size(30, 30) );
+
+        for ( size_t j = 0; j < nestedObjects.size(); j++ )
+        {
+            Rect nr = nestedObjects[j];
+            center.x = cvRound((r.x + nr.x + nr.width*0.5)*scale);
+            center.y = cvRound((r.y + nr.y + nr.height*0.5)*scale);
+            radius = cvRound((nr.width + nr.height)*0.25*scale);
+            circle( canvas, center, radius, color, 3, 8, 0 );
+        }
+    }
+    imshow( "result", canvas );
+}

modules/xobjdetect/src/cascadeclassifier.h

@@ -0,0 +1,95 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#ifndef _OPENCV_CASCADECLASSIFIER_H_
+#define _OPENCV_CASCADECLASSIFIER_H_
+
+#include "precomp.hpp"
+
+#define CC_CASCADE_FILENAME "cascade.xml"
+#define CC_PARAMS_FILENAME "params.xml"
+
+#define CC_CASCADE_PARAMS "cascadeParams"
+#define CC_STAGE_TYPE "stageType"
+#define CC_FEATURE_TYPE "featureType"
+#define CC_HEIGHT "height"
+#define CC_WIDTH  "width"
+
+#define CC_STAGE_NUM    "stageNum"
+#define CC_STAGES       "stages"
+#define CC_STAGE_PARAMS "stageParams"
+
+#define CC_BOOST            "BOOST"
+#define CC_BOOST_TYPE       "boostType"
+#define CC_DISCRETE_BOOST   "DAB"
+#define CC_REAL_BOOST       "RAB"
+#define CC_LOGIT_BOOST      "LB"
+#define CC_GENTLE_BOOST     "GAB"
+#define CC_MINHITRATE       "minHitRate"
+#define CC_MAXFALSEALARM    "maxFalseAlarm"
+#define CC_TRIM_RATE        "weightTrimRate"
+#define CC_MAX_DEPTH        "maxDepth"
+#define CC_WEAK_COUNT       "maxWeakCount"
+#define CC_STAGE_THRESHOLD  "stageThreshold"
+#define CC_WEAK_CLASSIFIERS "weakClassifiers"
+#define CC_INTERNAL_NODES   "internalNodes"
+#define CC_LEAF_VALUES      "leafValues"
+
+#define CC_FEATURES       FEATURES
+#define CC_FEATURE_PARAMS "featureParams"
+#define CC_MAX_CAT_COUNT  "maxCatCount"
+#define CC_FEATURE_SIZE   "featSize"
+
+#define CC_HAAR        "HAAR"
+#define CC_MODE        "mode"
+#define CC_MODE_BASIC  "BASIC"
+#define CC_MODE_CORE   "CORE"
+#define CC_MODE_ALL    "ALL"
+#define CC_RECTS       "rects"
+#define CC_TILTED      "tilted"
+
+#define CC_LBP  "LBP"
+#define CC_RECT "rect"
+
+#endif

modules/xobjdetect/src/cascadedetect.hpp

@@ -0,0 +1,656 @@
+#pragma once
+
+#include "opencv2/core/ocl.hpp"
+
+namespace cv
+{
+
+void clipObjects(Size sz, std::vector<Rect>& objects,
+                 std::vector<int>* a, std::vector<double>* b);
+
+class FeatureEvaluator
+{
+public:
+    enum
+    {
+        HAAR = 0,
+        LBP  = 1,
+        HOG  = 2
+    };
+
+    struct ScaleData
+    {
+        ScaleData() { scale = 0.f; layer_ofs = ystep = 0; }
+        Size getWorkingSize(Size winSize) const
+        {
+            return Size(std::max(szi.width - winSize.width, 0),
+                        std::max(szi.height - winSize.height, 0));
+        }
+
+        float scale;
+        Size szi;
+        int layer_ofs, ystep;
+    };
+
+    virtual ~FeatureEvaluator();
+
+    virtual bool read(const FileNode& node, Size origWinSize);
+    virtual Ptr<FeatureEvaluator> clone() const;
+    virtual int getFeatureType() const;
+    int getNumChannels() const { return nchannels; }
+
+    virtual bool setImage(InputArray img, const std::vector<float>& scales);
+    virtual bool setWindow(Point p, int scaleIdx);
+    const ScaleData& getScaleData(int scaleIdx) const
+    {
+        CV_Assert( 0 <= scaleIdx && scaleIdx < (int)scaleData->size());
+        return scaleData->at(scaleIdx);
+    }
+    virtual void getUMats(std::vector<UMat>& bufs);
+    virtual void getMats();
+
+    Size getLocalSize() const { return localSize; }
+    Size getLocalBufSize() const { return lbufSize; }
+
+    virtual float calcOrd(int featureIdx) const;
+    virtual int calcCat(int featureIdx) const;
+
+    static Ptr<FeatureEvaluator> create(int type);
+
+protected:
+    enum { SBUF_VALID=1, USBUF_VALID=2 };
+    int sbufFlag;
+
+    bool updateScaleData( Size imgsz, const std::vector<float>& _scales );
+    virtual void computeChannels( int, InputArray ) {}
+    virtual void computeOptFeatures() {}
+
+    Size origWinSize, sbufSize, localSize, lbufSize;
+    int nchannels;
+    Mat sbuf, rbuf;
+    UMat urbuf, usbuf, ufbuf, uscaleData;
+
+    Ptr<std::vector<ScaleData> > scaleData;
+};
+
+
+class CascadeClassifierImpl CV_FINAL : public BaseCascadeClassifier
+{
+public:
+    CascadeClassifierImpl();
+    virtual ~CascadeClassifierImpl() CV_OVERRIDE;
+
+    bool empty() const CV_OVERRIDE;
+    bool load( const String& filename ) CV_OVERRIDE;
+    void read( const FileNode& node ) CV_OVERRIDE;
+    bool read_( const FileNode& node );
+    void detectMultiScale( InputArray image,
+                          CV_OUT std::vector<Rect>& objects,
+                          double scaleFactor = 1.1,
+                          int minNeighbors = 3, int flags = 0,
+                          Size minSize = Size(),
+                          Size maxSize = Size() ) CV_OVERRIDE;
+
+    void detectMultiScale( InputArray image,
+                          CV_OUT std::vector<Rect>& objects,
+                          CV_OUT std::vector<int>& numDetections,
+                          double scaleFactor=1.1,
+                          int minNeighbors=3, int flags=0,
+                          Size minSize=Size(),
+                          Size maxSize=Size() ) CV_OVERRIDE;
+
+    void detectMultiScale( InputArray image,
+                          CV_OUT std::vector<Rect>& objects,
+                          CV_OUT std::vector<int>& rejectLevels,
+                          CV_OUT std::vector<double>& levelWeights,
+                          double scaleFactor = 1.1,
+                          int minNeighbors = 3, int flags = 0,
+                          Size minSize = Size(),
+                          Size maxSize = Size(),
+                          bool outputRejectLevels = false ) CV_OVERRIDE;
+
+
+    bool isOldFormatCascade() const CV_OVERRIDE;
+    Size getOriginalWindowSize() const CV_OVERRIDE;
+    int getFeatureType() const CV_OVERRIDE;
+    void* getOldCascade() CV_OVERRIDE;
+
+    void setMaskGenerator(const Ptr<MaskGenerator>& maskGenerator) CV_OVERRIDE;
+    Ptr<MaskGenerator> getMaskGenerator() CV_OVERRIDE;
+
+protected:
+    enum { SUM_ALIGN = 64 };
+
+    bool detectSingleScale( InputArray image, Size processingRectSize,
+                            int yStep, double factor, std::vector<Rect>& candidates,
+                            std::vector<int>& rejectLevels, std::vector<double>& levelWeights,
+                            Size sumSize0, bool outputRejectLevels = false );
+#ifdef HAVE_OPENCL
+    bool ocl_detectMultiScaleNoGrouping( const std::vector<float>& scales,
+                                         std::vector<Rect>& candidates );
+#endif
+    void detectMultiScaleNoGrouping( InputArray image, std::vector<Rect>& candidates,
+                                    std::vector<int>& rejectLevels, std::vector<double>& levelWeights,
+                                    double scaleFactor, Size minObjectSize, Size maxObjectSize,
+                                    bool outputRejectLevels = false );
+
+    enum { MAX_FACES = 10000 };
+    enum { BOOST = 0 };
+    enum { DO_CANNY_PRUNING    = CASCADE_DO_CANNY_PRUNING,
+        SCALE_IMAGE         = CASCADE_SCALE_IMAGE,
+        FIND_BIGGEST_OBJECT = CASCADE_FIND_BIGGEST_OBJECT,
+        DO_ROUGH_SEARCH     = CASCADE_DO_ROUGH_SEARCH
+    };
+
+    friend class CascadeClassifierInvoker;
+    friend class SparseCascadeClassifierInvoker;
+
+    template<class FEval>
+    friend int predictOrdered( CascadeClassifierImpl& cascade, Ptr<FeatureEvaluator> &featureEvaluator, double& weight);
+
+    template<class FEval>
+    friend int predictCategorical( CascadeClassifierImpl& cascade, Ptr<FeatureEvaluator> &featureEvaluator, double& weight);
+
+    template<class FEval>
+    friend int predictOrderedStump( CascadeClassifierImpl& cascade, Ptr<FeatureEvaluator> &featureEvaluator, double& weight);
+
+    template<class FEval>
+    friend int predictCategoricalStump( CascadeClassifierImpl& cascade, Ptr<FeatureEvaluator> &featureEvaluator, double& weight);
+
+    int runAt( Ptr<FeatureEvaluator>& feval, Point pt, int scaleIdx, double& weight );
+
+    class Data
+    {
+    public:
+        struct DTreeNode
+        {
+            int featureIdx;
+            float threshold; // for ordered features only
+            int left;
+            int right;
+        };
+
+        struct DTree
+        {
+            int nodeCount;
+        };
+
+        struct Stage
+        {
+            int first;
+            int ntrees;
+            float threshold;
+        };
+
+        struct Stump
+        {
+            Stump() : featureIdx(0), threshold(0), left(0), right(0) { }
+            Stump(int _featureIdx, float _threshold, float _left, float _right)
+            : featureIdx(_featureIdx), threshold(_threshold), left(_left), right(_right) {}
+
+            int featureIdx;
+            float threshold;
+            float left;
+            float right;
+        };
+
+        Data();
+
+        bool read(const FileNode &node);
+
+        int stageType;
+        int featureType;
+        int ncategories;
+        int minNodesPerTree, maxNodesPerTree;
+        Size origWinSize;
+
+        std::vector<Stage> stages;
+        std::vector<DTree> classifiers;
+        std::vector<DTreeNode> nodes;
+        std::vector<float> leaves;
+        std::vector<int> subsets;
+        std::vector<Stump> stumps;
+    };
+
+    Data data;
+    Ptr<FeatureEvaluator> featureEvaluator;
+    Ptr<CvHaarClassifierCascade> oldCascade;
+
+    Ptr<MaskGenerator> maskGenerator;
+    UMat ugrayImage;
+    UMat ufacepos, ustages, unodes, uleaves, usubsets;
+#ifdef HAVE_OPENCL
+    ocl::Kernel haarKernel, lbpKernel;
+    bool tryOpenCL;
+#endif
+
+    Mutex mtx;
+};
+
+#define CC_CASCADE_PARAMS "cascadeParams"
+#define CC_STAGE_TYPE     "stageType"
+#define CC_FEATURE_TYPE   "featureType"
+#define CC_HEIGHT         "height"
+#define CC_WIDTH          "width"
+
+#define CC_STAGE_NUM    "stageNum"
+#define CC_STAGES       "stages"
+#define CC_STAGE_PARAMS "stageParams"
+
+#define CC_BOOST            "BOOST"
+#define CC_MAX_DEPTH        "maxDepth"
+#define CC_WEAK_COUNT       "maxWeakCount"
+#define CC_STAGE_THRESHOLD  "stageThreshold"
+#define CC_WEAK_CLASSIFIERS "weakClassifiers"
+#define CC_INTERNAL_NODES   "internalNodes"
+#define CC_LEAF_VALUES      "leafValues"
+
+#define CC_FEATURES       "features"
+#define CC_FEATURE_PARAMS "featureParams"
+#define CC_MAX_CAT_COUNT  "maxCatCount"
+
+#define CC_HAAR   "HAAR"
+#define CC_RECTS  "rects"
+#define CC_TILTED "tilted"
+
+#define CC_LBP  "LBP"
+#define CC_RECT "rect"
+
+#define CC_HOG  "HOG"
+
+#define CV_SUM_PTRS( p0, p1, p2, p3, sum, rect, step )                    \
+    /* (x, y) */                                                          \
+    (p0) = sum + (rect).x + (step) * (rect).y,                            \
+    /* (x + w, y) */                                                      \
+    (p1) = sum + (rect).x + (rect).width + (step) * (rect).y,             \
+    /* (x, y + h) */                                                      \
+    (p2) = sum + (rect).x + (step) * ((rect).y + (rect).height),          \
+    /* (x + w, y + h) */                                                  \
+    (p3) = sum + (rect).x + (rect).width + (step) * ((rect).y + (rect).height)
+
+#define CV_TILTED_PTRS( p0, p1, p2, p3, tilted, rect, step )                        \
+    /* (x, y) */                                                                    \
+    (p0) = tilted + (rect).x + (step) * (rect).y,                                   \
+    /* (x - h, y + h) */                                                            \
+    (p1) = tilted + (rect).x - (rect).height + (step) * ((rect).y + (rect).height), \
+    /* (x + w, y + w) */                                                            \
+    (p2) = tilted + (rect).x + (rect).width + (step) * ((rect).y + (rect).width),   \
+    /* (x + w - h, y + w + h) */                                                    \
+    (p3) = tilted + (rect).x + (rect).width - (rect).height                         \
+           + (step) * ((rect).y + (rect).width + (rect).height)
+
+#define CALC_SUM_(p0, p1, p2, p3, offset) \
+    ((p0)[offset] - (p1)[offset] - (p2)[offset] + (p3)[offset])
+
+#define CALC_SUM(rect,offset) CALC_SUM_((rect)[0], (rect)[1], (rect)[2], (rect)[3], offset)
+
+#define CV_SUM_OFS( p0, p1, p2, p3, sum, rect, step )                 \
+/* (x, y) */                                                          \
+(p0) = sum + (rect).x + (step) * (rect).y,                            \
+/* (x + w, y) */                                                      \
+(p1) = sum + (rect).x + (rect).width + (step) * (rect).y,             \
+/* (x, y + h) */                                                      \
+(p2) = sum + (rect).x + (step) * ((rect).y + (rect).height),          \
+/* (x + w, y + h) */                                                  \
+(p3) = sum + (rect).x + (rect).width + (step) * ((rect).y + (rect).height)
+
+#define CV_TILTED_OFS( p0, p1, p2, p3, tilted, rect, step )                     \
+/* (x, y) */                                                                    \
+(p0) = tilted + (rect).x + (step) * (rect).y,                                   \
+/* (x - h, y + h) */                                                            \
+(p1) = tilted + (rect).x - (rect).height + (step) * ((rect).y + (rect).height), \
+/* (x + w, y + w) */                                                            \
+(p2) = tilted + (rect).x + (rect).width + (step) * ((rect).y + (rect).width),   \
+/* (x + w - h, y + w + h) */                                                    \
+(p3) = tilted + (rect).x + (rect).width - (rect).height                         \
++ (step) * ((rect).y + (rect).width + (rect).height)
+
+#define CALC_SUM_OFS_(p0, p1, p2, p3, ptr) \
+((ptr)[p0] - (ptr)[p1] - (ptr)[p2] + (ptr)[p3])
+
+#define CALC_SUM_OFS(rect, ptr) CALC_SUM_OFS_((rect)[0], (rect)[1], (rect)[2], (rect)[3], ptr)
+
+//----------------------------------------------  HaarEvaluator ---------------------------------------
+class HaarEvaluator CV_FINAL : public FeatureEvaluator
+{
+public:
+    struct Feature
+    {
+        Feature();
+        bool read(const FileNode& node, const Size& origWinSize);
+
+        bool tilted;
+
+        enum { RECT_NUM = 3 };
+        struct RectWeigth
+        {
+            Rect r;
+            float weight;
+        } rect[RECT_NUM];
+    };
+
+    struct OptFeature
+    {
+        OptFeature();
+
+        enum { RECT_NUM = Feature::RECT_NUM };
+        float calc( const int* pwin ) const;
+        void setOffsets( const Feature& _f, int step, int tofs );
+
+        int ofs[RECT_NUM][4];
+        float weight[4];
+    };
+
+    HaarEvaluator();
+    virtual ~HaarEvaluator() CV_OVERRIDE;
+
+    virtual bool read( const FileNode& node, Size origWinSize) CV_OVERRIDE;
+    virtual Ptr<FeatureEvaluator> clone() const CV_OVERRIDE;
+    virtual int getFeatureType() const CV_OVERRIDE { return FeatureEvaluator::HAAR; }
+
+    virtual bool setWindow(Point p, int scaleIdx) CV_OVERRIDE;
+    Rect getNormRect() const;
+    int getSquaresOffset() const;
+
+    float operator()(int featureIdx) const
+    { return optfeaturesPtr[featureIdx].calc(pwin) * varianceNormFactor; }
+    virtual float calcOrd(int featureIdx) const CV_OVERRIDE
+    { return (*this)(featureIdx); }
+
+protected:
+    virtual void computeChannels( int i, InputArray img ) CV_OVERRIDE;
+    virtual void computeOptFeatures() CV_OVERRIDE;
+
+    Ptr<std::vector<Feature> > features;
+    Ptr<std::vector<OptFeature> > optfeatures;
+    Ptr<std::vector<OptFeature> > optfeatures_lbuf;
+    bool hasTiltedFeatures;
+
+    int tofs, sqofs;
+    Vec4i nofs;
+    Rect normrect;
+    const int* pwin;
+    OptFeature* optfeaturesPtr; // optimization
+    float varianceNormFactor;
+};
+
+inline HaarEvaluator::Feature :: Feature()
+{
+    tilted = false;
+    rect[0].r = rect[1].r = rect[2].r = Rect();
+    rect[0].weight = rect[1].weight = rect[2].weight = 0;
+}
+
+inline HaarEvaluator::OptFeature :: OptFeature()
+{
+    weight[0] = weight[1] = weight[2] = 0.f;
+
+    ofs[0][0] = ofs[0][1] = ofs[0][2] = ofs[0][3] =
+    ofs[1][0] = ofs[1][1] = ofs[1][2] = ofs[1][3] =
+    ofs[2][0] = ofs[2][1] = ofs[2][2] = ofs[2][3] = 0;
+}
+
+inline float HaarEvaluator::OptFeature :: calc( const int* ptr ) const
+{
+    float ret = weight[0] * CALC_SUM_OFS(ofs[0], ptr) +
+                weight[1] * CALC_SUM_OFS(ofs[1], ptr);
+
+    if( weight[2] != 0.0f )
+        ret += weight[2] * CALC_SUM_OFS(ofs[2], ptr);
+
+    return ret;
+}
+
+//----------------------------------------------  LBPEvaluator -------------------------------------
+
+class LBPEvaluator CV_FINAL : public FeatureEvaluator
+{
+public:
+    struct Feature
+    {
+        Feature();
+        Feature( int x, int y, int _block_w, int _block_h  ) :
+                 rect(x, y, _block_w, _block_h) {}
+
+        bool read(const FileNode& node, const Size& origWinSize);
+
+        Rect rect; // weight and height for block
+    };
+
+    struct OptFeature
+    {
+        OptFeature();
+
+        int calc( const int* pwin ) const;
+        void setOffsets( const Feature& _f, int step );
+        int ofs[16];
+    };
+
+    LBPEvaluator();
+    virtual ~LBPEvaluator() CV_OVERRIDE;
+
+    virtual bool read( const FileNode& node, Size origWinSize ) CV_OVERRIDE;
+    virtual Ptr<FeatureEvaluator> clone() const CV_OVERRIDE;
+    virtual int getFeatureType() const CV_OVERRIDE { return FeatureEvaluator::LBP; }
+
+    virtual bool setWindow(Point p, int scaleIdx) CV_OVERRIDE;
+
+    int operator()(int featureIdx) const
+    { return optfeaturesPtr[featureIdx].calc(pwin); }
+    virtual int calcCat(int featureIdx) const CV_OVERRIDE
+    { return (*this)(featureIdx); }
+protected:
+    virtual void computeChannels( int i, InputArray img ) CV_OVERRIDE;
+    virtual void computeOptFeatures() CV_OVERRIDE;
+
+    Ptr<std::vector<Feature> > features;
+    Ptr<std::vector<OptFeature> > optfeatures;
+    Ptr<std::vector<OptFeature> > optfeatures_lbuf;
+    OptFeature* optfeaturesPtr; // optimization
+
+    const int* pwin;
+};
+
+
+inline LBPEvaluator::Feature :: Feature()
+{
+    rect = Rect();
+}
+
+inline LBPEvaluator::OptFeature :: OptFeature()
+{
+    for( int i = 0; i < 16; i++ )
+        ofs[i] = 0;
+}
+
+inline int LBPEvaluator::OptFeature :: calc( const int* p ) const
+{
+    int cval = CALC_SUM_OFS_( ofs[5], ofs[6], ofs[9], ofs[10], p );
+
+    return (CALC_SUM_OFS_( ofs[0], ofs[1], ofs[4], ofs[5], p ) >= cval ? 128 : 0) |   // 0
+           (CALC_SUM_OFS_( ofs[1], ofs[2], ofs[5], ofs[6], p ) >= cval ? 64 : 0) |    // 1
+           (CALC_SUM_OFS_( ofs[2], ofs[3], ofs[6], ofs[7], p ) >= cval ? 32 : 0) |    // 2
+           (CALC_SUM_OFS_( ofs[6], ofs[7], ofs[10], ofs[11], p ) >= cval ? 16 : 0) |  // 5
+           (CALC_SUM_OFS_( ofs[10], ofs[11], ofs[14], ofs[15], p ) >= cval ? 8 : 0)|  // 8
+           (CALC_SUM_OFS_( ofs[9], ofs[10], ofs[13], ofs[14], p ) >= cval ? 4 : 0)|   // 7
+           (CALC_SUM_OFS_( ofs[8], ofs[9], ofs[12], ofs[13], p ) >= cval ? 2 : 0)|    // 6
+           (CALC_SUM_OFS_( ofs[4], ofs[5], ofs[8], ofs[9], p ) >= cval ? 1 : 0);
+}
+
+
+//----------------------------------------------  predictor functions -------------------------------------
+
+template<class FEval>
+inline int predictOrdered( CascadeClassifierImpl& cascade,
+                           Ptr<FeatureEvaluator> &_featureEvaluator, double& sum )
+{
+    CV_INSTRUMENT_REGION();
+
+    int nstages = (int)cascade.data.stages.size();
+    int nodeOfs = 0, leafOfs = 0;
+    FEval& featureEvaluator = (FEval&)*_featureEvaluator;
+    float* cascadeLeaves = &cascade.data.leaves[0];
+    CascadeClassifierImpl::Data::DTreeNode* cascadeNodes = &cascade.data.nodes[0];
+    CascadeClassifierImpl::Data::DTree* cascadeWeaks = &cascade.data.classifiers[0];
+    CascadeClassifierImpl::Data::Stage* cascadeStages = &cascade.data.stages[0];
+
+    for( int si = 0; si < nstages; si++ )
+    {
+        CascadeClassifierImpl::Data::Stage& stage = cascadeStages[si];
+        int wi, ntrees = stage.ntrees;
+        sum = 0;
+
+        for( wi = 0; wi < ntrees; wi++ )
+        {
+            CascadeClassifierImpl::Data::DTree& weak = cascadeWeaks[stage.first + wi];
+            int idx = 0, root = nodeOfs;
+
+            do
+            {
+                CascadeClassifierImpl::Data::DTreeNode& node = cascadeNodes[root + idx];
+                double val = featureEvaluator(node.featureIdx);
+                idx = val < node.threshold ? node.left : node.right;
+            }
+            while( idx > 0 );
+            sum += cascadeLeaves[leafOfs - idx];
+            nodeOfs += weak.nodeCount;
+            leafOfs += weak.nodeCount + 1;
+        }
+        if( sum < stage.threshold )
+            return -si;
+    }
+    return 1;
+}
+
+template<class FEval>
+inline int predictCategorical( CascadeClassifierImpl& cascade,
+                               Ptr<FeatureEvaluator> &_featureEvaluator, double& sum )
+{
+    CV_INSTRUMENT_REGION();
+
+    int nstages = (int)cascade.data.stages.size();
+    int nodeOfs = 0, leafOfs = 0;
+    FEval& featureEvaluator = (FEval&)*_featureEvaluator;
+    size_t subsetSize = (cascade.data.ncategories + 31)/32;
+    int* cascadeSubsets = &cascade.data.subsets[0];
+    float* cascadeLeaves = &cascade.data.leaves[0];
+    CascadeClassifierImpl::Data::DTreeNode* cascadeNodes = &cascade.data.nodes[0];
+    CascadeClassifierImpl::Data::DTree* cascadeWeaks = &cascade.data.classifiers[0];
+    CascadeClassifierImpl::Data::Stage* cascadeStages = &cascade.data.stages[0];
+
+    for(int si = 0; si < nstages; si++ )
+    {
+        CascadeClassifierImpl::Data::Stage& stage = cascadeStages[si];
+        int wi, ntrees = stage.ntrees;
+        sum = 0;
+
+        for( wi = 0; wi < ntrees; wi++ )
+        {
+            CascadeClassifierImpl::Data::DTree& weak = cascadeWeaks[stage.first + wi];
+            int idx = 0, root = nodeOfs;
+            do
+            {
+                CascadeClassifierImpl::Data::DTreeNode& node = cascadeNodes[root + idx];
+                int c = featureEvaluator(node.featureIdx);
+                const int* subset = &cascadeSubsets[(root + idx)*subsetSize];
+                idx = (subset[c>>5] & (1 << (c & 31))) ? node.left : node.right;
+            }
+            while( idx > 0 );
+            sum += cascadeLeaves[leafOfs - idx];
+            nodeOfs += weak.nodeCount;
+            leafOfs += weak.nodeCount + 1;
+        }
+        if( sum < stage.threshold )
+            return -si;
+    }
+    return 1;
+}
+
+template<class FEval>
+inline int predictOrderedStump( CascadeClassifierImpl& cascade,
+                                Ptr<FeatureEvaluator> &_featureEvaluator, double& sum )
+{
+    CV_INSTRUMENT_REGION();
+
+    CV_Assert(!cascade.data.stumps.empty());
+    FEval& featureEvaluator = (FEval&)*_featureEvaluator;
+    const CascadeClassifierImpl::Data::Stump* cascadeStumps = &cascade.data.stumps[0];
+    const CascadeClassifierImpl::Data::Stage* cascadeStages = &cascade.data.stages[0];
+
+    int nstages = (int)cascade.data.stages.size();
+    double tmp = 0;
+
+    for( int stageIdx = 0; stageIdx < nstages; stageIdx++ )
+    {
+        const CascadeClassifierImpl::Data::Stage& stage = cascadeStages[stageIdx];
+        tmp = 0;
+
+        int ntrees = stage.ntrees;
+        for( int i = 0; i < ntrees; i++ )
+        {
+            const CascadeClassifierImpl::Data::Stump& stump = cascadeStumps[i];
+            double value = featureEvaluator(stump.featureIdx);
+            tmp += value < stump.threshold ? stump.left : stump.right;
+        }
+
+        if( tmp < stage.threshold )
+        {
+            sum = (double)tmp;
+            return -stageIdx;
+        }
+        cascadeStumps += ntrees;
+    }
+
+    sum = (double)tmp;
+    return 1;
+}
+
+template<class FEval>
+inline int predictCategoricalStump( CascadeClassifierImpl& cascade,
+                                    Ptr<FeatureEvaluator> &_featureEvaluator, double& sum )
+{
+    CV_INSTRUMENT_REGION();
+
+    CV_Assert(!cascade.data.stumps.empty());
+    int nstages = (int)cascade.data.stages.size();
+    FEval& featureEvaluator = (FEval&)*_featureEvaluator;
+    size_t subsetSize = (cascade.data.ncategories + 31)/32;
+    const int* cascadeSubsets = &cascade.data.subsets[0];
+    const CascadeClassifierImpl::Data::Stump* cascadeStumps = &cascade.data.stumps[0];
+    const CascadeClassifierImpl::Data::Stage* cascadeStages = &cascade.data.stages[0];
+
+    double tmp = 0;
+    for( int si = 0; si < nstages; si++ )
+    {
+        const CascadeClassifierImpl::Data::Stage& stage = cascadeStages[si];
+        int wi, ntrees = stage.ntrees;
+        tmp = 0;
+
+        for( wi = 0; wi < ntrees; wi++ )
+        {
+            const CascadeClassifierImpl::Data::Stump& stump = cascadeStumps[wi];
+            int c = featureEvaluator(stump.featureIdx);
+            const int* subset = &cascadeSubsets[wi*subsetSize];
+            tmp += (subset[c>>5] & (1 << (c & 31))) ? stump.left : stump.right;
+        }
+
+        if( tmp < stage.threshold )
+        {
+            sum = tmp;
+            return -si;
+        }
+
+        cascadeStumps += ntrees;
+        cascadeSubsets += ntrees*subsetSize;
+    }
+
+    sum = (double)tmp;
+    return 1;
+}
+
+namespace haar_cvt
+{
+bool convert(const FileNode& oldcascade_root, FileStorage& newfs);
+}
+
+}

modules/xobjdetect/src/cascadedetect_convert.cpp

@@ -0,0 +1,275 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2013, Itseez Inc, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of Intel Corporation may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+/* Haar features calculation */
+
+#include "precomp.hpp"
+#include "cascadedetect.hpp"
+#include <stdio.h>
+
+namespace cv
+{
+
+/* field names */
+
+#define ICV_HAAR_SIZE_NAME            "size"
+#define ICV_HAAR_STAGES_NAME          "stages"
+#define ICV_HAAR_TREES_NAME           "trees"
+#define ICV_HAAR_FEATURE_NAME         "feature"
+#define ICV_HAAR_RECTS_NAME           "rects"
+#define ICV_HAAR_TILTED_NAME          "tilted"
+#define ICV_HAAR_THRESHOLD_NAME       "threshold"
+#define ICV_HAAR_LEFT_NODE_NAME       "left_node"
+#define ICV_HAAR_LEFT_VAL_NAME        "left_val"
+#define ICV_HAAR_RIGHT_NODE_NAME      "right_node"
+#define ICV_HAAR_RIGHT_VAL_NAME       "right_val"
+#define ICV_HAAR_STAGE_THRESHOLD_NAME "stage_threshold"
+#define ICV_HAAR_PARENT_NAME          "parent"
+#define ICV_HAAR_NEXT_NAME            "next"
+
+namespace haar_cvt
+{
+
+struct HaarFeature
+{
+    enum { RECT_NUM = 3 };
+
+    HaarFeature()
+    {
+        tilted = false;
+        for( int i = 0; i < RECT_NUM; i++ )
+        {
+            rect[i].r = Rect(0,0,0,0);
+            rect[i].weight = 0.f;
+        }
+    }
+    bool tilted;
+    struct
+    {
+        Rect r;
+        float weight;
+    } rect[RECT_NUM];
+};
+
+struct HaarClassifierNode
+{
+    HaarClassifierNode()
+    {
+        f = left = right = 0;
+        threshold = 0.f;
+    }
+    int f, left, right;
+    float threshold;
+};
+
+struct HaarClassifier
+{
+    std::vector<HaarClassifierNode> nodes;
+    std::vector<float> leaves;
+};
+
+struct HaarStageClassifier
+{
+    HaarStageClassifier() : threshold(0) {}
+
+    double threshold;
+    std::vector<HaarClassifier> weaks;
+};
+
+bool convert(const FileNode& oldroot, FileStorage& newfs)
+{
+    FileNode sznode = oldroot[ICV_HAAR_SIZE_NAME];
+    if( sznode.empty() )
+        return false;
+    Size cascadesize;
+    cascadesize.width = (int)sznode[0];
+    cascadesize.height = (int)sznode[1];
+    std::vector<HaarFeature> features;
+
+    int i, j, k, n;
+
+    FileNode stages_seq = oldroot[ICV_HAAR_STAGES_NAME];
+    int nstages = (int)stages_seq.size();
+    std::vector<HaarStageClassifier> stages(nstages);
+
+    for( i = 0; i < nstages; i++ )
+    {
+        FileNode stagenode = stages_seq[i];
+        HaarStageClassifier& stage = stages[i];
+        stage.threshold = (double)stagenode[ICV_HAAR_STAGE_THRESHOLD_NAME];
+        FileNode weaks_seq = stagenode[ICV_HAAR_TREES_NAME];
+        int nweaks = (int)weaks_seq.size();
+        stage.weaks.resize(nweaks);
+
+        for( j = 0; j < nweaks; j++ )
+        {
+            HaarClassifier& weak = stage.weaks[j];
+            FileNode weaknode = weaks_seq[j];
+            int nnodes = (int)weaknode.size();
+
+            for( n = 0; n < nnodes; n++ )
+            {
+                FileNode nnode = weaknode[n];
+                FileNode fnode = nnode[ICV_HAAR_FEATURE_NAME];
+                HaarFeature f;
+                HaarClassifierNode node;
+                node.f = (int)features.size();
+                f.tilted = (int)fnode[ICV_HAAR_TILTED_NAME] != 0;
+                FileNode rects_seq = fnode[ICV_HAAR_RECTS_NAME];
+                int nrects = (int)rects_seq.size();
+
+                for( k = 0; k < nrects; k++ )
+                {
+                    FileNode rnode = rects_seq[k];
+                    f.rect[k].r.x = (int)rnode[0];
+                    f.rect[k].r.y = (int)rnode[1];
+                    f.rect[k].r.width = (int)rnode[2];
+                    f.rect[k].r.height = (int)rnode[3];
+                    f.rect[k].weight = (float)rnode[4];
+                }
+                features.push_back(f);
+                node.threshold = nnode[ICV_HAAR_THRESHOLD_NAME];
+                FileNode leftValNode = nnode[ICV_HAAR_LEFT_VAL_NAME];
+                if( !leftValNode.empty() )
+                {
+                    node.left = -(int)weak.leaves.size();
+                    weak.leaves.push_back((float)leftValNode);
+                }
+                else
+                {
+                    node.left = (int)nnode[ICV_HAAR_LEFT_NODE_NAME];
+                }
+                FileNode rightValNode = nnode[ICV_HAAR_RIGHT_VAL_NAME];
+                if( !rightValNode.empty() )
+                {
+                    node.right = -(int)weak.leaves.size();
+                    weak.leaves.push_back((float)rightValNode);
+                }
+                else
+                {
+                    node.right = (int)nnode[ICV_HAAR_RIGHT_NODE_NAME];
+                }
+                weak.nodes.push_back(node);
+            }
+        }
+    }
+
+    int maxWeakCount = 0, nfeatures = (int)features.size();
+    for( i = 0; i < nstages; i++ )
+        maxWeakCount = std::max(maxWeakCount, (int)stages[i].weaks.size());
+
+    newfs << "cascade" << "{:opencv-cascade-classifier"
+    << "stageType" << "BOOST"
+    << "featureType" << "HAAR"
+    << "width" << cascadesize.width
+    << "height" << cascadesize.height
+    << "stageParams" << "{"
+        << "maxWeakCount" << (int)maxWeakCount
+    << "}"
+    << "featureParams" << "{"
+        << "maxCatCount" << 0
+    << "}"
+    << "stageNum" << (int)nstages
+    << "stages" << "[";
+
+    for( i = 0; i < nstages; i++ )
+    {
+        int nweaks = (int)stages[i].weaks.size();
+        newfs << "{" << "maxWeakCount" << (int)nweaks
+            << "stageThreshold" << stages[i].threshold
+            << "weakClassifiers" << "[";
+        for( j = 0; j < nweaks; j++ )
+        {
+            const HaarClassifier& c = stages[i].weaks[j];
+            newfs << "{" << "internalNodes" << "[:";
+            int nnodes = (int)c.nodes.size(), nleaves = (int)c.leaves.size();
+            for( k = 0; k < nnodes; k++ )
+                newfs << c.nodes[k].left << c.nodes[k].right
+                    << c.nodes[k].f << c.nodes[k].threshold;
+            newfs << "]" << "leafValues" << "[:";
+            for( k = 0; k < nleaves; k++ )
+                newfs << c.leaves[k];
+            newfs << "]" << "}";
+        }
+        newfs << "]" << "}";
+    }
+
+    newfs << "]"
+        << "features" << "[";
+
+    for( i = 0; i < nfeatures; i++ )
+    {
+        const HaarFeature& f = features[i];
+        newfs << "{" << "rects" << "[";
+        for( j = 0; j < HaarFeature::RECT_NUM; j++ )
+        {
+            if( j >= 2 && fabs(f.rect[j].weight) < FLT_EPSILON )
+                break;
+            newfs << "[:" << f.rect[j].r.x << f.rect[j].r.y <<
+                f.rect[j].r.width << f.rect[j].r.height << f.rect[j].weight << "]";
+        }
+        newfs << "]";
+        if( f.tilted )
+            newfs << "tilted" << 1;
+        newfs << "}";
+    }
+
+    newfs << "]" << "}";
+    return true;
+}
+
+}
+
+bool CascadeClassifier::convert(const String& oldcascade, const String& newcascade)
+{
+    FileStorage oldfs(oldcascade, FileStorage::READ);
+    FileStorage newfs(newcascade, FileStorage::WRITE);
+    if( !oldfs.isOpened() || !newfs.isOpened() )
+        return false;
+    FileNode oldroot = oldfs.getFirstTopLevelNode();
+
+    bool ok = haar_cvt::convert(oldroot, newfs);
+    if( !ok && newcascade.size() > 0 )
+        remove(newcascade.c_str());
+    return ok;
+}
+
+}

modules/xobjdetect/src/detection_based_tracker.cpp

@@ -0,0 +1,885 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "precomp.hpp"
+
+#include "opencv2/core/utility.hpp"
+
+#include <thread>
+#include <mutex>
+#include <condition_variable>
+
+#if defined(DEBUG) || defined(_DEBUG)
+#undef DEBUGLOGS
+#define DEBUGLOGS 1
+#endif
+
+#ifndef DEBUGLOGS
+#define DEBUGLOGS 0
+#endif
+
+#ifdef __ANDROID__
+#include <android/log.h>
+#define LOG_TAG "OBJECT_DETECTOR"
+#define LOGD0(...) ((void)__android_log_print(ANDROID_LOG_DEBUG, LOG_TAG, __VA_ARGS__))
+#define LOGI0(...) ((void)__android_log_print(ANDROID_LOG_INFO, LOG_TAG, __VA_ARGS__))
+#define LOGW0(...) ((void)__android_log_print(ANDROID_LOG_WARN, LOG_TAG, __VA_ARGS__))
+#define LOGE0(...) ((void)__android_log_print(ANDROID_LOG_ERROR, LOG_TAG, __VA_ARGS__))
+#else
+
+#include <stdio.h>
+
+#define LOGD0(_str, ...) (printf(_str , ## __VA_ARGS__), printf("\n"), fflush(stdout))
+#define LOGI0(_str, ...) (printf(_str , ## __VA_ARGS__), printf("\n"), fflush(stdout))
+#define LOGW0(_str, ...) (printf(_str , ## __VA_ARGS__), printf("\n"), fflush(stdout))
+#define LOGE0(_str, ...) (printf(_str , ## __VA_ARGS__), printf("\n"), fflush(stdout))
+#endif //__ANDROID__
+
+#if DEBUGLOGS
+#define LOGD(_str, ...) LOGD0(_str , ## __VA_ARGS__)
+#define LOGI(_str, ...) LOGI0(_str , ## __VA_ARGS__)
+#define LOGW(_str, ...) LOGW0(_str , ## __VA_ARGS__)
+#define LOGE(_str, ...) LOGE0(_str , ## __VA_ARGS__)
+#else
+#define LOGD(...)
+#define LOGI(...)
+#define LOGW(...)
+#define LOGE(...)
+#endif //DEBUGLOGS
+
+
+using namespace cv;
+
+static inline cv::Point2f centerRect(const cv::Rect& r)
+{
+    return cv::Point2f(r.x+((float)r.width)/2, r.y+((float)r.height)/2);
+}
+
+static inline cv::Rect scale_rect(const cv::Rect& r, float scale)
+{
+    cv::Point2f m=centerRect(r);
+    float width  = r.width  * scale;
+    float height = r.height * scale;
+    int x=cvRound(m.x - width/2);
+    int y=cvRound(m.y - height/2);
+
+    return cv::Rect(x, y, cvRound(width), cvRound(height));
+}
+
+namespace cv
+{
+    void* workcycleObjectDetectorFunction(void* p);
+}
+
+class cv::DetectionBasedTracker::SeparateDetectionWork
+{
+    public:
+        SeparateDetectionWork(cv::DetectionBasedTracker& _detectionBasedTracker, cv::Ptr<DetectionBasedTracker::IDetector> _detector,
+                              const cv::DetectionBasedTracker::Parameters& params);
+        virtual ~SeparateDetectionWork();
+        bool communicateWithDetectingThread(const Mat& imageGray, std::vector<Rect>& rectsWhereRegions);
+        bool run();
+        void stop();
+        void resetTracking();
+
+        inline bool isWorking()
+        {
+            return (stateThread==STATE_THREAD_WORKING_SLEEPING) || (stateThread==STATE_THREAD_WORKING_WITH_IMAGE);
+        }
+        void setParameters(const cv::DetectionBasedTracker::Parameters& params)
+        {
+            std::unique_lock<std::mutex> mtx_lock(mtx);
+            parameters = params;
+        }
+
+        inline void init()
+        {
+            std::unique_lock<std::mutex> mtx_lock(mtx);
+            stateThread = STATE_THREAD_STOPPED;
+            isObjectDetectingReady = false;
+            shouldObjectDetectingResultsBeForgot = false;
+            objectDetectorThreadStartStop.notify_one();
+        }
+    protected:
+
+        DetectionBasedTracker& detectionBasedTracker;
+        cv::Ptr<DetectionBasedTracker::IDetector> cascadeInThread;
+        std::thread second_workthread;
+        std::mutex mtx;
+        std::condition_variable objectDetectorRun;
+        std::condition_variable objectDetectorThreadStartStop;
+        std::vector<cv::Rect> resultDetect;
+        volatile bool isObjectDetectingReady;
+        volatile bool shouldObjectDetectingResultsBeForgot;
+
+        enum StateSeparatedThread {
+            STATE_THREAD_STOPPED=0,
+            STATE_THREAD_WORKING_SLEEPING,
+            STATE_THREAD_WORKING_WITH_IMAGE,
+            STATE_THREAD_WORKING,
+            STATE_THREAD_STOPPING
+        };
+        volatile StateSeparatedThread stateThread;
+
+        cv::Mat imageSeparateDetecting;
+
+        void workcycleObjectDetector();
+        friend void* workcycleObjectDetectorFunction(void* p);
+
+        long long  timeWhenDetectingThreadStartedWork;
+        cv::DetectionBasedTracker::Parameters parameters;
+};
+
+cv::DetectionBasedTracker::SeparateDetectionWork::SeparateDetectionWork(DetectionBasedTracker& _detectionBasedTracker, cv::Ptr<DetectionBasedTracker::IDetector> _detector,
+                                                                        const cv::DetectionBasedTracker::Parameters& params)
+    :detectionBasedTracker(_detectionBasedTracker),
+    cascadeInThread(),
+    isObjectDetectingReady(false),
+    shouldObjectDetectingResultsBeForgot(false),
+    stateThread(STATE_THREAD_STOPPED),
+    timeWhenDetectingThreadStartedWork(-1),
+    parameters(params)
+{
+    CV_Assert(_detector);
+
+    cascadeInThread = _detector;
+}
+
+cv::DetectionBasedTracker::SeparateDetectionWork::~SeparateDetectionWork()
+{
+    if(stateThread!=STATE_THREAD_STOPPED) {
+        LOGE("\n\n\nATTENTION!!! dangerous algorithm error: destructor DetectionBasedTracker::DetectionBasedTracker::~SeparateDetectionWork is called before stopping the workthread");
+    }
+    second_workthread.join();
+}
+bool cv::DetectionBasedTracker::SeparateDetectionWork::run()
+{
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::run() --- start");
+    std::unique_lock<std::mutex> mtx_lock(mtx);
+    // unlocked when leaving scope
+    if (stateThread != STATE_THREAD_STOPPED) {
+        LOGE("DetectionBasedTracker::SeparateDetectionWork::run is called while the previous run is not stopped");
+        return false;
+    }
+    stateThread=STATE_THREAD_WORKING_SLEEPING;
+    second_workthread = std::thread(workcycleObjectDetectorFunction, (void*)this); //TODO: add attributes?
+    objectDetectorThreadStartStop.wait(mtx_lock);
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::run --- end");
+    return true;
+}
+
+#define CATCH_ALL_AND_LOG(_block)                                                           \
+    try {                                                                                   \
+        _block;                                                                             \
+    }                                                                                       \
+    catch(const cv::Exception& e) {                                                         \
+        LOGE0("\n %s: ERROR: OpenCV Exception caught: \n'%s'\n\n", CV_Func, e.what());      \
+    } catch(const std::exception& e) {                                                      \
+        LOGE0("\n %s: ERROR: Exception caught: \n'%s'\n\n", CV_Func, e.what());             \
+    } catch(...) {                                                                          \
+        LOGE0("\n %s: ERROR: UNKNOWN Exception caught\n\n", CV_Func);                       \
+    }
+
+void* cv::workcycleObjectDetectorFunction(void* p)
+{
+    CATCH_ALL_AND_LOG({ ((cv::DetectionBasedTracker::SeparateDetectionWork*)p)->workcycleObjectDetector(); });
+    try{
+        ((cv::DetectionBasedTracker::SeparateDetectionWork*)p)->init();
+    } catch(...) {
+        LOGE0("DetectionBasedTracker: workcycleObjectDetectorFunction: ERROR concerning pointer, received as the function parameter");
+    }
+    return NULL;
+}
+
+void cv::DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector()
+{
+    static double freq = getTickFrequency();
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- start");
+    std::vector<Rect> objects;
+
+    CV_Assert(stateThread==STATE_THREAD_WORKING_SLEEPING);
+    std::unique_lock<std::mutex> mtx_lock(mtx);
+    {
+        objectDetectorThreadStartStop.notify_one();
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- before waiting");
+        CV_Assert(stateThread==STATE_THREAD_WORKING_SLEEPING);
+        objectDetectorRun.wait(mtx_lock);
+        if (isWorking()) {
+            stateThread=STATE_THREAD_WORKING_WITH_IMAGE;
+        }
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- after waiting");
+    }
+    mtx_lock.unlock();
+
+    bool isFirstStep=true;
+
+    isObjectDetectingReady=false;
+
+    while(isWorking())
+    {
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- next step");
+
+        if (! isFirstStep) {
+            LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- before waiting");
+            CV_Assert(stateThread==STATE_THREAD_WORKING_SLEEPING);
+            mtx_lock.lock();
+            if (!isWorking()) {//it is a rare case, but may cause a crash
+                LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- go out from the workcycle from inner part of lock just before waiting");
+                mtx_lock.unlock();
+                break;
+            }
+            CV_Assert(stateThread==STATE_THREAD_WORKING_SLEEPING);
+            objectDetectorRun.wait(mtx_lock);
+            if (isWorking()) {
+                stateThread=STATE_THREAD_WORKING_WITH_IMAGE;
+            }
+            mtx_lock.unlock();
+
+            LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- after waiting");
+        } else {
+            isFirstStep=false;
+        }
+
+        if (!isWorking()) {
+            LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- go out from the workcycle just after waiting");
+            break;
+        }
+
+
+        if (imageSeparateDetecting.empty()) {
+            LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- imageSeparateDetecting is empty, continue");
+            continue;
+        }
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- start handling imageSeparateDetecting, img.size=%dx%d, img.data=0x%p",
+                imageSeparateDetecting.size().width, imageSeparateDetecting.size().height, (void*)imageSeparateDetecting.data);
+
+
+        int64 t1_detect=getTickCount();
+
+        cascadeInThread->detect(imageSeparateDetecting, objects);
+
+        /*cascadeInThread.detectMultiScale( imageSeparateDetecting, objects,
+                detectionBasedTracker.parameters.scaleFactor, detectionBasedTracker.parameters.minNeighbors, 0
+                |CV_HAAR_SCALE_IMAGE
+                ,
+                min_objectSize,
+                max_objectSize
+                );
+        */
+
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- end handling imageSeparateDetecting");
+
+        if (!isWorking()) {
+            LOGD("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- go out from the workcycle just after detecting");
+            break;
+        }
+
+        int64 t2_detect = getTickCount();
+        int64 dt_detect = t2_detect-t1_detect;
+        double dt_detect_ms=((double)dt_detect)/freq * 1000.0;
+        (void)(dt_detect_ms);
+
+        LOGI("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector() --- objects num==%d, t_ms=%.4f", (int)objects.size(), dt_detect_ms);
+        mtx_lock.lock();
+        if (!shouldObjectDetectingResultsBeForgot) {
+            resultDetect=objects;
+            isObjectDetectingReady=true;
+        } else { //shouldObjectDetectingResultsBeForgot==true
+            resultDetect.clear();
+            isObjectDetectingReady=false;
+            shouldObjectDetectingResultsBeForgot=false;
+        }
+        if(isWorking()) {
+            stateThread=STATE_THREAD_WORKING_SLEEPING;
+        }
+        mtx_lock.unlock();
+
+        objects.clear();
+    }// while(isWorking())
+
+    LOGI("DetectionBasedTracker::SeparateDetectionWork::workcycleObjectDetector: Returning");
+}
+
+void cv::DetectionBasedTracker::SeparateDetectionWork::stop()
+{
+    //FIXME: TODO: should add quickStop functionality
+  std::unique_lock<std::mutex> mtx_lock(mtx);
+    if (!isWorking()) {
+        mtx_lock.unlock();
+        LOGE("SimpleHighguiDemoCore::stop is called but the SimpleHighguiDemoCore pthread is not active");
+        stateThread = STATE_THREAD_STOPPING;
+        return;
+    }
+    stateThread=STATE_THREAD_STOPPING;
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::stop: before going to sleep to wait for the signal from the workthread");
+    objectDetectorRun.notify_one();
+    objectDetectorThreadStartStop.wait(mtx_lock);
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::stop: after receiving the signal from the workthread, stateThread=%d", (int)stateThread);
+    mtx_lock.unlock();
+}
+
+void cv::DetectionBasedTracker::SeparateDetectionWork::resetTracking()
+{
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::resetTracking");
+    std::unique_lock<std::mutex> mtx_lock(mtx);
+
+    if (stateThread == STATE_THREAD_WORKING_WITH_IMAGE) {
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::resetTracking: since workthread is detecting objects at the moment, we should make cascadeInThread stop detecting and forget the detecting results");
+        shouldObjectDetectingResultsBeForgot=true;
+        //cascadeInThread.setStopFlag();//FIXME: TODO: this feature also should be contributed to OpenCV
+    } else {
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::resetTracking: since workthread is NOT detecting objects at the moment, we should NOT make any additional actions");
+    }
+
+    resultDetect.clear();
+    isObjectDetectingReady=false;
+
+    mtx_lock.unlock();
+}
+
+bool cv::DetectionBasedTracker::SeparateDetectionWork::communicateWithDetectingThread(const Mat& imageGray, std::vector<Rect>& rectsWhereRegions)
+{
+    static double freq = getTickFrequency();
+
+    bool shouldCommunicateWithDetectingThread = (stateThread==STATE_THREAD_WORKING_SLEEPING);
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::communicateWithDetectingThread: shouldCommunicateWithDetectingThread=%d", (shouldCommunicateWithDetectingThread?1:0));
+
+    if (!shouldCommunicateWithDetectingThread) {
+        return false;
+    }
+
+    bool shouldHandleResult = false;
+
+    std::unique_lock<std::mutex> mtx_lock(mtx);
+
+    if (isObjectDetectingReady) {
+        shouldHandleResult=true;
+        rectsWhereRegions = resultDetect;
+        isObjectDetectingReady=false;
+
+        double lastBigDetectionDuration = 1000.0 * (((double)(getTickCount()  - timeWhenDetectingThreadStartedWork )) / freq);
+        (void)(lastBigDetectionDuration);
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::communicateWithDetectingThread: lastBigDetectionDuration=%f ms", (double)lastBigDetectionDuration);
+    }
+
+    bool shouldSendNewDataToWorkThread = true;
+    if (timeWhenDetectingThreadStartedWork > 0) {
+        double time_from_previous_launch_in_ms=1000.0 * (((double)(getTickCount()  - timeWhenDetectingThreadStartedWork )) / freq); //the same formula as for lastBigDetectionDuration
+        shouldSendNewDataToWorkThread = (time_from_previous_launch_in_ms >= detectionBasedTracker.parameters.minDetectionPeriod);
+        LOGD("DetectionBasedTracker::SeparateDetectionWork::communicateWithDetectingThread: shouldSendNewDataToWorkThread was 1, now it is %d, since time_from_previous_launch_in_ms=%.2f, minDetectionPeriod=%d",
+                (shouldSendNewDataToWorkThread?1:0), time_from_previous_launch_in_ms, detectionBasedTracker.parameters.minDetectionPeriod);
+    }
+
+    if (shouldSendNewDataToWorkThread) {
+
+        imageSeparateDetecting.create(imageGray.size(), CV_8UC1);
+
+        imageGray.copyTo(imageSeparateDetecting);//may change imageSeparateDetecting ptr. But should not.
+
+
+        timeWhenDetectingThreadStartedWork = getTickCount() ;
+
+        objectDetectorRun.notify_one();
+    }
+
+    mtx_lock.unlock();
+    LOGD("DetectionBasedTracker::SeparateDetectionWork::communicateWithDetectingThread: result: shouldHandleResult=%d", (shouldHandleResult?1:0));
+
+    return shouldHandleResult;
+}
+
+cv::DetectionBasedTracker::Parameters::Parameters()
+{
+  maxTrackLifetime = 5;
+  minDetectionPeriod = 0;
+}
+
+cv::DetectionBasedTracker::InnerParameters::InnerParameters()
+{
+    numLastPositionsToTrack=4;
+    numStepsToWaitBeforeFirstShow=6;
+    numStepsToTrackWithoutDetectingIfObjectHasNotBeenShown=3;
+    numStepsToShowWithoutDetecting=3;
+
+    coeffTrackingWindowSize=2.0;
+    coeffObjectSizeToTrack=0.85f;
+    coeffObjectSpeedUsingInPrediction=0.8f;
+
+}
+
+cv::DetectionBasedTracker::DetectionBasedTracker(cv::Ptr<IDetector> mainDetector, cv::Ptr<IDetector> trackingDetector, const Parameters& params)
+    :separateDetectionWork(),
+    parameters(params),
+    innerParameters(),
+    numTrackedSteps(0),
+    cascadeForTracking(trackingDetector)
+{
+    CV_Assert( (params.maxTrackLifetime >= 0)
+//            && mainDetector
+            && trackingDetector );
+
+    if (mainDetector) {
+        Ptr<SeparateDetectionWork> tmp(new SeparateDetectionWork(*this, mainDetector, params));
+        separateDetectionWork.swap(tmp);
+    }
+
+    weightsPositionsSmoothing.push_back(1);
+    weightsSizesSmoothing.push_back(0.5);
+    weightsSizesSmoothing.push_back(0.3f);
+    weightsSizesSmoothing.push_back(0.2f);
+}
+
+cv::DetectionBasedTracker::~DetectionBasedTracker()
+{
+}
+
+void DetectionBasedTracker::process(const Mat& imageGray)
+{
+    CV_INSTRUMENT_REGION();
+
+    CV_Assert(imageGray.type()==CV_8UC1);
+
+    if ( separateDetectionWork && !separateDetectionWork->isWorking() ) {
+        separateDetectionWork->run();
+    }
+
+    static double freq = getTickFrequency();
+    static long long time_when_last_call_started=getTickCount();
+
+    {
+        double delta_time_from_prev_call=1000.0 * (((double)(getTickCount()  - time_when_last_call_started)) / freq);
+        (void)(delta_time_from_prev_call);
+        LOGD("DetectionBasedTracker::process: time from the previous call is %f ms", (double)delta_time_from_prev_call);
+        time_when_last_call_started=getTickCount();
+    }
+
+    Mat imageDetect=imageGray;
+
+    std::vector<Rect> rectsWhereRegions;
+    bool shouldHandleResult=false;
+    if (separateDetectionWork) {
+        shouldHandleResult = separateDetectionWork->communicateWithDetectingThread(imageGray, rectsWhereRegions);
+    }
+
+    if (shouldHandleResult) {
+        LOGD("DetectionBasedTracker::process: get _rectsWhereRegions were got from resultDetect");
+    } else {
+        LOGD("DetectionBasedTracker::process: get _rectsWhereRegions from previous positions");
+        for(size_t i = 0; i < trackedObjects.size(); i++) {
+            size_t n = trackedObjects[i].lastPositions.size();
+            CV_Assert(n > 0);
+
+            Rect r = trackedObjects[i].lastPositions[n-1];
+            if(r.empty()) {
+                LOGE("DetectionBasedTracker::process: ERROR: ATTENTION: strange algorithm's behavior: trackedObjects[i].rect() is empty");
+                continue;
+            }
+
+            //correction by speed of rectangle
+            if (n > 1) {
+                Point2f center = centerRect(r);
+                Point2f center_prev = centerRect(trackedObjects[i].lastPositions[n-2]);
+                Point2f shift = (center - center_prev) * innerParameters.coeffObjectSpeedUsingInPrediction;
+
+                r.x += cvRound(shift.x);
+                r.y += cvRound(shift.y);
+            }
+
+
+            rectsWhereRegions.push_back(r);
+        }
+    }
+    LOGI("DetectionBasedTracker::process: tracked objects num==%d", (int)trackedObjects.size());
+
+    std::vector<Rect> detectedObjectsInRegions;
+
+    LOGD("DetectionBasedTracker::process: rectsWhereRegions.size()=%d", (int)rectsWhereRegions.size());
+    for(size_t i=0; i < rectsWhereRegions.size(); i++) {
+        Rect r = rectsWhereRegions[i];
+
+        detectInRegion(imageDetect, r, detectedObjectsInRegions);
+    }
+    LOGD("DetectionBasedTracker::process: detectedObjectsInRegions.size()=%d", (int)detectedObjectsInRegions.size());
+
+    updateTrackedObjects(detectedObjectsInRegions);
+}
+
+void cv::DetectionBasedTracker::getObjects(std::vector<cv::Rect>& result) const
+{
+    result.clear();
+
+    for(size_t i=0; i < trackedObjects.size(); i++) {
+        Rect r=calcTrackedObjectPositionToShow((int)i);
+        if (r.empty()) {
+            continue;
+        }
+        result.push_back(r);
+        LOGD("DetectionBasedTracker::process: found a object with SIZE %d x %d, rect={%d, %d, %d x %d}", r.width, r.height, r.x, r.y, r.width, r.height);
+    }
+}
+
+void cv::DetectionBasedTracker::getObjects(std::vector<Object>& result) const
+{
+    result.clear();
+
+    for(size_t i=0; i < trackedObjects.size(); i++) {
+        Rect r=calcTrackedObjectPositionToShow((int)i);
+        if (r.empty()) {
+            continue;
+        }
+        result.push_back(Object(r, trackedObjects[i].id));
+        LOGD("DetectionBasedTracker::process: found a object with SIZE %d x %d, rect={%d, %d, %d x %d}", r.width, r.height, r.x, r.y, r.width, r.height);
+    }
+}
+void cv::DetectionBasedTracker::getObjects(std::vector<ExtObject>& result) const
+{
+    result.clear();
+
+    for(size_t i=0; i < trackedObjects.size(); i++) {
+        ObjectStatus status;
+        Rect r=calcTrackedObjectPositionToShow((int)i, status);
+        result.push_back(ExtObject(trackedObjects[i].id, r, status));
+        LOGD("DetectionBasedTracker::process: found a object with SIZE %d x %d, rect={%d, %d, %d x %d}, status = %d", r.width, r.height, r.x, r.y, r.width, r.height, (int)status);
+    }
+}
+
+bool cv::DetectionBasedTracker::run()
+{
+    if (separateDetectionWork) {
+        return separateDetectionWork->run();
+    }
+    return false;
+}
+
+void cv::DetectionBasedTracker::stop()
+{
+    if (separateDetectionWork) {
+        separateDetectionWork->stop();
+    }
+}
+
+void cv::DetectionBasedTracker::resetTracking()
+{
+    if (separateDetectionWork) {
+        separateDetectionWork->resetTracking();
+    }
+    trackedObjects.clear();
+}
+
+void cv::DetectionBasedTracker::updateTrackedObjects(const std::vector<Rect>& detectedObjects)
+{
+    enum {
+        NEW_RECTANGLE=-1,
+        INTERSECTED_RECTANGLE=-2
+    };
+
+    int N1=(int)trackedObjects.size();
+    int N2=(int)detectedObjects.size();
+    LOGD("DetectionBasedTracker::updateTrackedObjects: N1=%d, N2=%d", N1, N2);
+
+    for(int i=0; i < N1; i++) {
+        trackedObjects[i].numDetectedFrames++;
+    }
+
+    std::vector<int> correspondence(detectedObjects.size(), NEW_RECTANGLE);
+    correspondence.clear();
+    correspondence.resize(detectedObjects.size(), NEW_RECTANGLE);
+
+    for(int i=0; i < N1; i++) {
+        LOGD("DetectionBasedTracker::updateTrackedObjects: i=%d", i);
+        TrackedObject& curObject=trackedObjects[i];
+
+        int bestIndex=-1;
+        int bestArea=-1;
+
+        int numpositions=(int)curObject.lastPositions.size();
+        CV_Assert(numpositions > 0);
+        Rect prevRect=curObject.lastPositions[numpositions-1];
+        LOGD("DetectionBasedTracker::updateTrackedObjects: prevRect[%d]={%d, %d, %d x %d}", i, prevRect.x, prevRect.y, prevRect.width, prevRect.height);
+
+        for(int j=0; j < N2; j++) {
+            LOGD("DetectionBasedTracker::updateTrackedObjects: j=%d", j);
+            if (correspondence[j] >= 0) {
+                LOGD("DetectionBasedTracker::updateTrackedObjects: j=%d is rejected, because it has correspondence=%d", j, correspondence[j]);
+                continue;
+            }
+            if (correspondence[j] !=NEW_RECTANGLE) {
+                LOGD("DetectionBasedTracker::updateTrackedObjects: j=%d is rejected, because it is intersected with another rectangle", j);
+                continue;
+            }
+            LOGD("DetectionBasedTracker::updateTrackedObjects: detectedObjects[%d]={%d, %d, %d x %d}",
+                    j, detectedObjects[j].x, detectedObjects[j].y, detectedObjects[j].width, detectedObjects[j].height);
+
+            Rect r=prevRect & detectedObjects[j];
+            if ( (r.width > 0) && (r.height > 0) ) {
+                LOGD("DetectionBasedTracker::updateTrackedObjects: There is intersection between prevRect and detectedRect, r={%d, %d, %d x %d}",
+                        r.x, r.y, r.width, r.height);
+                correspondence[j]=INTERSECTED_RECTANGLE;
+
+                if ( r.area() > bestArea) {
+                    LOGD("DetectionBasedTracker::updateTrackedObjects: The area of intersection is %d, it is better than bestArea=%d", r.area(), bestArea);
+                    bestIndex=j;
+                    bestArea=r.area();
+                }
+            }
+        }
+        if (bestIndex >= 0) {
+            LOGD("DetectionBasedTracker::updateTrackedObjects: The best correspondence for i=%d is j=%d", i, bestIndex);
+            correspondence[bestIndex]=i;
+
+            for(int j=0; j < N2; j++) {
+                if (correspondence[j] >= 0)
+                    continue;
+
+                Rect r=detectedObjects[j] & detectedObjects[bestIndex];
+                if ( (r.width > 0) && (r.height > 0) ) {
+                    LOGD("DetectionBasedTracker::updateTrackedObjects: Found intersection between "
+                            "rectangles j=%d and bestIndex=%d, rectangle j=%d is marked as intersected", j, bestIndex, j);
+                    correspondence[j]=INTERSECTED_RECTANGLE;
+                }
+            }
+        } else {
+            LOGD("DetectionBasedTracker::updateTrackedObjects: There is no correspondence for i=%d ", i);
+            curObject.numFramesNotDetected++;
+        }
+    }
+
+    LOGD("DetectionBasedTracker::updateTrackedObjects: start second cycle");
+    for(int j=0; j < N2; j++) {
+        LOGD("DetectionBasedTracker::updateTrackedObjects: j=%d", j);
+        int i=correspondence[j];
+        if (i >= 0) {//add position
+            LOGD("DetectionBasedTracker::updateTrackedObjects: add position");
+            trackedObjects[i].lastPositions.push_back(detectedObjects[j]);
+            while ((int)trackedObjects[i].lastPositions.size() > (int) innerParameters.numLastPositionsToTrack) {
+                trackedObjects[i].lastPositions.erase(trackedObjects[i].lastPositions.begin());
+            }
+            trackedObjects[i].numFramesNotDetected=0;
+        } else if (i==NEW_RECTANGLE){ //new object
+            LOGD("DetectionBasedTracker::updateTrackedObjects: new object");
+            trackedObjects.push_back(detectedObjects[j]);
+        } else {
+            LOGD("DetectionBasedTracker::updateTrackedObjects: was auxiliary intersection");
+        }
+    }
+
+    std::vector<TrackedObject>::iterator it=trackedObjects.begin();
+    while( it != trackedObjects.end() ) {
+        if ( (it->numFramesNotDetected > parameters.maxTrackLifetime)
+                ||
+                (
+                 (it->numDetectedFrames <= innerParameters.numStepsToWaitBeforeFirstShow)
+                 &&
+                 (it->numFramesNotDetected > innerParameters.numStepsToTrackWithoutDetectingIfObjectHasNotBeenShown)
+                )
+           )
+        {
+            int numpos=(int)it->lastPositions.size();
+            CV_Assert(numpos > 0);
+            Rect r = it->lastPositions[numpos-1];
+            (void)(r);
+            LOGD("DetectionBasedTracker::updateTrackedObjects: deleted object {%d, %d, %d x %d}",
+                    r.x, r.y, r.width, r.height);
+            it=trackedObjects.erase(it);
+        } else {
+            it++;
+        }
+    }
+}
+
+int cv::DetectionBasedTracker::addObject(const Rect& location)
+{
+    LOGD("DetectionBasedTracker::addObject: new object {%d, %d %dx%d}",location.x, location.y, location.width, location.height);
+    trackedObjects.push_back(TrackedObject(location));
+    int newId = trackedObjects.back().id;
+    LOGD("DetectionBasedTracker::addObject: newId = %d", newId);
+    return newId;
+}
+
+Rect cv::DetectionBasedTracker::calcTrackedObjectPositionToShow(int i) const
+{
+    ObjectStatus status;
+    return calcTrackedObjectPositionToShow(i, status);
+}
+Rect cv::DetectionBasedTracker::calcTrackedObjectPositionToShow(int i, ObjectStatus& status) const
+{
+    if ( (i < 0) || (i >= (int)trackedObjects.size()) ) {
+        LOGE("DetectionBasedTracker::calcTrackedObjectPositionToShow: ERROR: wrong i=%d", i);
+        status = WRONG_OBJECT;
+        return Rect();
+    }
+    if (trackedObjects[i].numDetectedFrames <= innerParameters.numStepsToWaitBeforeFirstShow){
+        LOGI("DetectionBasedTracker::calcTrackedObjectPositionToShow: trackedObjects[%d].numDetectedFrames=%d <= numStepsToWaitBeforeFirstShow=%d --- return empty Rect()",
+                i, trackedObjects[i].numDetectedFrames, innerParameters.numStepsToWaitBeforeFirstShow);
+        status = DETECTED_NOT_SHOWN_YET;
+        return Rect();
+    }
+    if (trackedObjects[i].numFramesNotDetected > innerParameters.numStepsToShowWithoutDetecting) {
+        status = DETECTED_TEMPORARY_LOST;
+        return Rect();
+    }
+
+    const TrackedObject::PositionsVector& lastPositions=trackedObjects[i].lastPositions;
+
+    int N=(int)lastPositions.size();
+    if (N<=0) {
+        LOGE("DetectionBasedTracker::calcTrackedObjectPositionToShow: ERROR: no positions for i=%d", i);
+        status = WRONG_OBJECT;
+        return Rect();
+    }
+
+    int Nsize=std::min(N, (int)weightsSizesSmoothing.size());
+    int Ncenter= std::min(N, (int)weightsPositionsSmoothing.size());
+
+    Point2f center;
+    double w=0, h=0;
+    if (Nsize > 0) {
+        double sum=0;
+        for(int j=0; j < Nsize; j++) {
+            int k=N-j-1;
+            w += lastPositions[k].width  * weightsSizesSmoothing[j];
+            h += lastPositions[k].height * weightsSizesSmoothing[j];
+            sum+=weightsSizesSmoothing[j];
+        }
+        w /= sum;
+        h /= sum;
+    } else {
+        w=lastPositions[N-1].width;
+        h=lastPositions[N-1].height;
+    }
+
+    if (Ncenter > 0) {
+        double sum=0;
+        for(int j=0; j < Ncenter; j++) {
+            int k=N-j-1;
+            Point tl(lastPositions[k].tl());
+            Point br(lastPositions[k].br());
+            Point2f c1;
+            c1=tl;
+            c1=c1* 0.5f;
+            Point2f c2;
+            c2=br;
+            c2=c2*0.5f;
+            c1=c1+c2;
+
+            center=center+  (c1  * weightsPositionsSmoothing[j]);
+            sum+=weightsPositionsSmoothing[j];
+        }
+        center *= (float)(1 / sum);
+    } else {
+        int k=N-1;
+        Point tl(lastPositions[k].tl());
+        Point br(lastPositions[k].br());
+        Point2f c1;
+        c1=tl;
+        c1=c1* 0.5f;
+        Point2f c2;
+        c2=br;
+        c2=c2*0.5f;
+
+        center=c1+c2;
+    }
+    Point2f tl=center-Point2f((float)w*0.5f,(float)h*0.5f);
+    Rect res(cvRound(tl.x), cvRound(tl.y), cvRound(w), cvRound(h));
+    LOGD("DetectionBasedTracker::calcTrackedObjectPositionToShow: Result for i=%d: {%d, %d, %d x %d}", i, res.x, res.y, res.width, res.height);
+
+    status = DETECTED;
+    return res;
+}
+
+void cv::DetectionBasedTracker::detectInRegion(const Mat& img, const Rect& r, std::vector<Rect>& detectedObjectsInRegions)
+{
+    Rect r0(Point(), img.size());
+    Rect r1 = scale_rect(r, innerParameters.coeffTrackingWindowSize);
+    r1 = r1 & r0;
+
+    if ( (r1.width <=0) || (r1.height <= 0) ) {
+        LOGD("DetectionBasedTracker::detectInRegion: Empty intersection");
+        return;
+    }
+
+    int d = cvRound(std::min(r.width, r.height) * innerParameters.coeffObjectSizeToTrack);
+
+    std::vector<Rect> tmpobjects;
+
+    Mat img1(img, r1);//subimage for rectangle -- without data copying
+    LOGD("DetectionBasedTracker::detectInRegion: img1.size()=%d x %d, d=%d",
+            img1.size().width, img1.size().height, d);
+
+    cascadeForTracking->setMinObjectSize(Size(d, d));
+    cascadeForTracking->detect(img1, tmpobjects);
+            /*
+            detectMultiScale( img1, tmpobjects,
+            parameters.scaleFactor, parameters.minNeighbors, 0
+            |CV_HAAR_FIND_BIGGEST_OBJECT
+            |CV_HAAR_SCALE_IMAGE
+            ,
+            Size(d,d),
+            max_objectSize
+            );*/
+
+    for(size_t i=0; i < tmpobjects.size(); i++) {
+        Rect curres(tmpobjects[i].tl() + r1.tl(), tmpobjects[i].size());
+        detectedObjectsInRegions.push_back(curres);
+    }
+}
+
+bool cv::DetectionBasedTracker::setParameters(const Parameters& params)
+{
+    if ( params.maxTrackLifetime < 0 )
+    {
+        LOGE("DetectionBasedTracker::setParameters: ERROR: wrong parameters value");
+        return false;
+    }
+
+    if (separateDetectionWork) {
+        separateDetectionWork->setParameters(params);
+    }
+    parameters=params;
+    return true;
+}
+
+const cv::DetectionBasedTracker::Parameters& DetectionBasedTracker::getParameters() const
+{
+    return parameters;
+}

modules/xobjdetect/src/feature_evaluator.cpp

@@ -0,0 +1,132 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#include "precomp.hpp"
+#include "cascadeclassifier.h"
+
+namespace cv {
+namespace xobjdetect {
+
+float calcNormFactor( const Mat& sum, const Mat& sqSum )
+{
+    CV_DbgAssert( sum.cols > 3 && sqSum.rows > 3 );
+    Rect normrect( 1, 1, sum.cols - 3, sum.rows - 3 );
+    size_t p0, p1, p2, p3;
+    CV_SUM_OFFSETS( p0, p1, p2, p3, normrect, sum.step1() )
+    double area = normrect.width * normrect.height;
+    const int *sp = sum.ptr<int>();
+    int valSum = sp[p0] - sp[p1] - sp[p2] + sp[p3];
+    const double *sqp = sqSum.ptr<double>();
+    double valSqSum = sqp[p0] - sqp[p1] - sqp[p2] + sqp[p3];
+    return (float) sqrt( (double) (area * valSqSum - (double)valSum * valSum) );
+}
+
+CvParams::CvParams() : name( "params" ) {}
+void CvParams::printDefaults() const
+{ std::cout << "--" << name << "--" << std::endl; }
+void CvParams::printAttrs() const {}
+bool CvParams::scanAttr( const std::string, const std::string ) { return false; }
+
+
+//---------------------------- FeatureParams --------------------------------------
+
+CvFeatureParams::CvFeatureParams() : maxCatCount( 0 ), featSize( 1 )
+{
+    name = CC_FEATURE_PARAMS;
+}
+
+void CvFeatureParams::init( const CvFeatureParams& fp )
+{
+    maxCatCount = fp.maxCatCount;
+    featSize = fp.featSize;
+}
+
+void CvFeatureParams::write( FileStorage &fs ) const
+{
+    fs << CC_MAX_CAT_COUNT << maxCatCount;
+    fs << CC_FEATURE_SIZE << featSize;
+}
+
+bool CvFeatureParams::read( const FileNode &node )
+{
+    if ( node.empty() )
+        return false;
+    maxCatCount = node[CC_MAX_CAT_COUNT];
+    featSize = node[CC_FEATURE_SIZE];
+    return ( maxCatCount >= 0 && featSize >= 1 );
+}
+
+Ptr<CvFeatureParams> CvFeatureParams::create()
+{
+    return Ptr<CvFeatureParams>(new CvLBPFeatureParams);
+}
+
+//------------------------------------- FeatureEvaluator ---------------------------------------
+
+void CvFeatureEvaluator::init(const CvFeatureParams *_featureParams,
+                              int _maxSampleCount, Size _winSize )
+{
+    CV_Assert(_maxSampleCount > 0);
+    featureParams = (CvFeatureParams *)_featureParams;
+    winSize = _winSize;
+    numFeatures = 0;
+    cls.create( (int)_maxSampleCount, 1, CV_32FC1 );
+    generateFeatures();
+}
+
+void CvFeatureEvaluator::setImage(const Mat &, uchar clsLabel, int idx, const std::vector<int>&)
+{
+    //CV_Assert(img.cols == winSize.width);
+    //CV_Assert(img.rows == winSize.height);
+    CV_Assert(idx < cls.rows);
+    cls.ptr<float>(idx)[0] = clsLabel;
+}
+
+Ptr<CvFeatureEvaluator> CvFeatureEvaluator::create()
+{
+    return Ptr<CvFeatureEvaluator>(new CvLBPEvaluator);
+}
+
+}
+}

modules/xobjdetect/src/feature_evaluator.hpp

@@ -0,0 +1,151 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#ifndef _OPENCV_FEATURES_H_
+#define _OPENCV_FEATURES_H_
+
+#include <stdio.h>
+
+#define FEATURES "features"
+
+#define CV_SUM_OFFSETS( p0, p1, p2, p3, rect, step )                      \
+    /* (x, y) */                                                          \
+    (p0) = (rect).x + (step) * (rect).y;                                  \
+    /* (x + w, y) */                                                      \
+    (p1) = (rect).x + (rect).width + (step) * (rect).y;                   \
+    /* (x + w, y) */                                                      \
+    (p2) = (rect).x + (step) * ((rect).y + (rect).height);                \
+    /* (x + w, y + h) */                                                  \
+    (p3) = (rect).x + (rect).width + (step) * ((rect).y + (rect).height);
+
+#define CV_TILTED_OFFSETS( p0, p1, p2, p3, rect, step )                   \
+    /* (x, y) */                                                          \
+    (p0) = (rect).x + (step) * (rect).y;                                  \
+    /* (x - h, y + h) */                                                  \
+    (p1) = (rect).x - (rect).height + (step) * ((rect).y + (rect).height);\
+    /* (x + w, y + w) */                                                  \
+    (p2) = (rect).x + (rect).width + (step) * ((rect).y + (rect).width);  \
+    /* (x + w - h, y + w + h) */                                          \
+    (p3) = (rect).x + (rect).width - (rect).height                        \
+           + (step) * ((rect).y + (rect).width + (rect).height);
+
+namespace cv {
+namespace xobjdetect {
+
+float calcNormFactor( const cv::Mat& sum, const cv::Mat& sqSum );
+
+template<class Feature>
+void _writeFeatures( const std::vector<Feature> features, cv::FileStorage &fs, const cv::Mat& featureMap )
+{
+    fs << FEATURES << "[";
+    const cv::Mat_<int>& featureMap_ = (const cv::Mat_<int>&)featureMap;
+    for ( int fi = 0; fi < featureMap.cols; fi++ )
+        if ( featureMap_(0, fi) >= 0 )
+        {
+            fs << "{";
+            features[fi].write( fs );
+            fs << "}";
+        }
+    fs << "]";
+}
+
+class CvParams
+{
+public:
+    CvParams();
+    virtual ~CvParams() {}
+    // from|to file
+    virtual void write( cv::FileStorage &fs ) const = 0;
+    virtual bool read( const cv::FileNode &node ) = 0;
+    // from|to screen
+    virtual void printDefaults() const;
+    virtual void printAttrs() const;
+    virtual bool scanAttr( const std::string prmName, const std::string val );
+    std::string name;
+};
+
+class CvFeatureParams : public CvParams
+{
+public:
+    enum { HAAR = 0, LBP = 1, HOG = 2 };
+    CvFeatureParams();
+    virtual void init( const CvFeatureParams& fp );
+    virtual void write( cv::FileStorage &fs ) const CV_OVERRIDE;
+    virtual bool read( const cv::FileNode &node ) CV_OVERRIDE;
+    static cv::Ptr<CvFeatureParams> create();
+    int maxCatCount; // 0 in case of numerical features
+    int featSize; // 1 in case of simple features (HAAR, LBP) and N_BINS(9)*N_CELLS(4) in case of Dalal's HOG features
+};
+
+class CvFeatureEvaluator
+{
+public:
+    virtual ~CvFeatureEvaluator() {}
+    virtual void init(const CvFeatureParams *_featureParams,
+                      int _maxSampleCount, cv::Size _winSize );
+    virtual void setImage(const cv::Mat& img, uchar clsLabel, int idx, const std::vector<int> &feature_ind);
+    virtual void setWindow(const cv::Point& p) = 0;
+    virtual void writeFeatures( cv::FileStorage &fs, const cv::Mat& featureMap ) const = 0;
+    virtual float operator()(int featureIdx) = 0;
+    static cv::Ptr<CvFeatureEvaluator> create();
+
+    int getNumFeatures() const { return numFeatures; }
+    int getMaxCatCount() const { return featureParams->maxCatCount; }
+    int getFeatureSize() const { return featureParams->featSize; }
+    const cv::Mat& getCls() const { return cls; }
+    float getCls(int si) const { return cls.at<float>(si, 0); }
+protected:
+    virtual void generateFeatures() = 0;
+
+    int npos, nneg;
+    int numFeatures;
+    cv::Size winSize;
+    CvFeatureParams *featureParams;
+    cv::Mat cls;
+};
+
+}
+}
+
+#endif

modules/xobjdetect/src/lbpfeatures.cpp

@@ -0,0 +1,129 @@
+
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#include "precomp.hpp"
+#include "cascadeclassifier.h"
+
+namespace cv {
+namespace xobjdetect {
+
+CvLBPFeatureParams::CvLBPFeatureParams()
+{
+    maxCatCount = 256;
+    name = LBPF_NAME;
+}
+
+void CvLBPEvaluator::init(const CvFeatureParams *_featureParams, int _maxSampleCount, Size _winSize)
+{
+    CV_Assert( _maxSampleCount > 0);
+    sum.create((int)_maxSampleCount, (_winSize.width + 1) * (_winSize.height + 1), CV_32SC1);
+    CvFeatureEvaluator::init( _featureParams, _maxSampleCount, _winSize );
+}
+
+void CvLBPEvaluator::setImage(const Mat &img, uchar clsLabel, int idx,
+    const std::vector<int> &feature_ind)
+{
+    CV_DbgAssert( !sum.empty() );
+    CvFeatureEvaluator::setImage( img, clsLabel, idx, feature_ind );
+    integral( img, sum );
+    cur_sum = sum;
+    offset_ = int(sum.ptr<int>(1) - sum.ptr<int>());
+    for (size_t i = 0; i < feature_ind.size(); ++i) {
+        features[feature_ind[i]].calcPoints(offset_);
+    }
+}
+
+void CvLBPEvaluator::writeFeatures( FileStorage &fs, const Mat& featureMap ) const
+{
+    _writeFeatures( features, fs, featureMap );
+}
+
+void CvLBPEvaluator::generateFeatures()
+{
+    int offset = winSize.width + 1;
+    for( int x = 0; x < winSize.width; x++ )
+        for( int y = 0; y < winSize.height; y++ )
+            for( int w = 1; w <= winSize.width / 3; w++ )
+                for( int h = 1; h <= winSize.height / 3; h++ )
+                    if ( (x+3*w <= winSize.width) && (y+3*h <= winSize.height) )
+                        features.push_back( Feature(offset, x, y, w, h ) );
+    numFeatures = (int)features.size();
+}
+
+CvLBPEvaluator::Feature::Feature()
+{
+    rect = Rect(0, 0, 0, 0);
+}
+
+CvLBPEvaluator::Feature::Feature( int offset, int x, int y, int _blockWidth, int _blockHeight )
+{
+    x_ = x;
+    y_ = y;
+    block_w_ = _blockWidth;
+    block_h_ = _blockHeight;
+    offset_ = offset;
+    calcPoints(offset);
+}
+
+void CvLBPEvaluator::Feature::calcPoints(int offset)
+{
+    Rect tr = rect = Rect(x_, y_, block_w_, block_h_);
+    CV_SUM_OFFSETS( p[0], p[1], p[4], p[5], tr, offset )
+    tr.x += 2*rect.width;
+    CV_SUM_OFFSETS( p[2], p[3], p[6], p[7], tr, offset )
+    tr.y +=2*rect.height;
+    CV_SUM_OFFSETS( p[10], p[11], p[14], p[15], tr, offset )
+    tr.x -= 2*rect.width;
+    CV_SUM_OFFSETS( p[8], p[9], p[12], p[13], tr, offset )
+    offset_ = offset;
+}
+
+void CvLBPEvaluator::Feature::write(FileStorage &fs) const
+{
+    fs << CC_RECT << "[:" << rect.x << rect.y << rect.width << rect.height << "]";
+}
+
+}
+}

modules/xobjdetect/src/lbpfeatures.h

@@ -0,0 +1,116 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#ifndef _OPENCV_LBPFEATURES_H_
+#define _OPENCV_LBPFEATURES_H_
+
+#include "precomp.hpp"
+
+
+#define LBPF_NAME "lbpFeatureParams"
+
+namespace cv {
+namespace xobjdetect {
+
+struct CvLBPFeatureParams : CvFeatureParams
+{
+    CvLBPFeatureParams();
+
+};
+
+class CvLBPEvaluator : public CvFeatureEvaluator
+{
+public:
+    virtual ~CvLBPEvaluator() {}
+    virtual void init(const CvFeatureParams *_featureParams,
+        int _maxSampleCount, cv::Size _winSize ) CV_OVERRIDE;
+    virtual void setImage(const cv::Mat& img, uchar clsLabel, int idx, const std::vector<int> &feature_ind) CV_OVERRIDE;
+    virtual void setWindow(const cv::Point& p) CV_OVERRIDE
+    { cur_sum = sum.rowRange(p.y, p.y + winSize.height).colRange(p.x, p.x + winSize.width); }
+    virtual float operator()(int featureIdx) CV_OVERRIDE
+    { return (float)features[featureIdx].calc( cur_sum ); }
+    virtual void writeFeatures( cv::FileStorage &fs, const cv::Mat& featureMap ) const CV_OVERRIDE;
+protected:
+    virtual void generateFeatures() CV_OVERRIDE;
+
+    class Feature
+    {
+    public:
+        Feature();
+        Feature( int offset, int x, int y, int _block_w, int _block_h  );
+        uchar calc( const cv::Mat& _sum );
+        void write( cv::FileStorage &fs ) const;
+
+        cv::Rect rect;
+        int p[16];
+
+        int x_, y_, block_w_, block_h_, offset_;
+        void calcPoints(int offset);
+    };
+    std::vector<Feature> features;
+
+    cv::Mat sum, cur_sum;
+    int offset_;
+};
+
+inline uchar CvLBPEvaluator::Feature::calc(const cv::Mat &_sum)
+{
+    const int* psum = _sum.ptr<int>();
+
+    int cval = psum[p[5]] - psum[p[6]] - psum[p[9]] + psum[p[10]];
+
+    return (uchar)((psum[p[0]] - psum[p[1]] - psum[p[4]] + psum[p[5]] >= cval ? 128 : 0) |   // 0
+        (psum[p[1]] - psum[p[2]] - psum[p[5]] + psum[p[6]] >= cval ? 64 : 0) |    // 1
+        (psum[p[2]] - psum[p[3]] - psum[p[6]] + psum[p[7]] >= cval ? 32 : 0) |    // 2
+        (psum[p[6]] - psum[p[7]] - psum[p[10]] + psum[p[11]] >= cval ? 16 : 0) |  // 5
+        (psum[p[10]] - psum[p[11]] - psum[p[14]] + psum[p[15]] >= cval ? 8 : 0) | // 8
+        (psum[p[9]] - psum[p[10]] - psum[p[13]] + psum[p[14]] >= cval ? 4 : 0) |  // 7
+        (psum[p[8]] - psum[p[9]] - psum[p[12]] + psum[p[13]] >= cval ? 2 : 0) |   // 6
+        (psum[p[4]] - psum[p[5]] - psum[p[8]] + psum[p[9]] >= cval ? 1 : 0));     // 3
+}
+
+}
+}
+
+#endif

modules/xobjdetect/src/opencl/cascadedetect.cl

@@ -0,0 +1,661 @@
+///////////////////////////// OpenCL kernels for face detection //////////////////////////////
+////////////////////////////// see the opencv/doc/license.txt ///////////////////////////////
+
+//
+// the code has been derived from the OpenCL Haar cascade kernel by
+//
+//    Niko Li, newlife20080214@gmail.com
+//    Wang Weiyan, wangweiyanster@gmail.com
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Nathan, liujun@multicorewareinc.com
+//    Peng Xiao, pengxiao@outlook.com
+//    Erping Pang, erping@multicorewareinc.com
+//
+
+#ifdef HAAR
+typedef struct __attribute__((aligned(4))) OptHaarFeature
+{
+    int4 ofs[3] __attribute__((aligned (4)));
+    float4 weight __attribute__((aligned (4)));
+}
+OptHaarFeature;
+#endif
+
+#ifdef LBP
+typedef struct __attribute__((aligned(4))) OptLBPFeature
+{
+    int16 ofs __attribute__((aligned (4)));
+}
+OptLBPFeature;
+#endif
+
+typedef struct __attribute__((aligned(4))) Stump
+{
+    float4 st __attribute__((aligned (4)));
+}
+Stump;
+
+typedef struct __attribute__((aligned(4))) Node
+{
+    int4 n __attribute__((aligned (4)));
+}
+Node;
+
+typedef struct __attribute__((aligned (4))) Stage
+{
+    int first __attribute__((aligned (4)));
+    int ntrees __attribute__((aligned (4)));
+    float threshold __attribute__((aligned (4)));
+}
+Stage;
+
+typedef struct __attribute__((aligned (4))) ScaleData
+{
+    float scale __attribute__((aligned (4)));
+    int szi_width __attribute__((aligned (4)));
+    int szi_height __attribute__((aligned (4)));
+    int layer_ofs __attribute__((aligned (4)));
+    int ystep __attribute__((aligned (4)));
+}
+ScaleData;
+
+#ifndef SUM_BUF_SIZE
+#define SUM_BUF_SIZE 0
+#endif
+
+#ifndef NODE_COUNT
+#define NODE_COUNT 1
+#endif
+
+#ifdef HAAR
+__kernel __attribute__((reqd_work_group_size(LOCAL_SIZE_X,LOCAL_SIZE_Y,1)))
+void runHaarClassifier(
+    int nscales, __global const ScaleData* scaleData,
+    __global const int* sum,
+    int _sumstep, int sumoffset,
+    __global const OptHaarFeature* optfeatures,
+    __global const Stage* stages,
+    __global const Node* nodes,
+    __global const float* leaves0,
+
+    volatile __global int* facepos,
+    int4 normrect, int sqofs, int2 windowsize)
+{
+    int lx = get_local_id(0);
+    int ly = get_local_id(1);
+    int groupIdx = get_group_id(0);
+    int i, ngroups = get_global_size(0)/LOCAL_SIZE_X;
+    int scaleIdx, tileIdx, stageIdx;
+    int sumstep = (int)(_sumstep/sizeof(int));
+    int4 nofs0 = (int4)(mad24(normrect.y, sumstep, normrect.x),
+                        mad24(normrect.y, sumstep, normrect.x + normrect.z),
+                        mad24(normrect.y + normrect.w, sumstep, normrect.x),
+                        mad24(normrect.y + normrect.w, sumstep, normrect.x + normrect.z));
+    int normarea = normrect.z * normrect.w;
+    float invarea = 1.f/normarea;
+    int lidx = ly*LOCAL_SIZE_X + lx;
+
+    #if SUM_BUF_SIZE > 0
+    int4 nofs = (int4)(mad24(normrect.y, SUM_BUF_STEP, normrect.x),
+                       mad24(normrect.y, SUM_BUF_STEP, normrect.x + normrect.z),
+                       mad24(normrect.y + normrect.w, SUM_BUF_STEP, normrect.x),
+                       mad24(normrect.y + normrect.w, SUM_BUF_STEP, normrect.x + normrect.z));
+    #else
+    int4 nofs = nofs0;
+    #endif
+    #define LOCAL_SIZE (LOCAL_SIZE_X*LOCAL_SIZE_Y)
+    __local int lstore[SUM_BUF_SIZE + LOCAL_SIZE*5/2+1];
+    #if SUM_BUF_SIZE > 0
+    __local int* ibuf = lstore;
+    __local int* lcount = ibuf + SUM_BUF_SIZE;
+    #else
+    __local int* lcount = lstore;
+    #endif
+    __local float* lnf = (__local float*)(lcount + 1);
+    __local float* lpartsum = lnf + LOCAL_SIZE;
+    __local short* lbuf = (__local short*)(lpartsum + LOCAL_SIZE);
+
+    for( scaleIdx = nscales-1; scaleIdx >= 0; scaleIdx-- )
+    {
+        __global const ScaleData* s = scaleData + scaleIdx;
+        int ystep = s->ystep;
+        int2 worksize = (int2)(max(s->szi_width - windowsize.x, 0), max(s->szi_height - windowsize.y, 0));
+        int2 ntiles = (int2)((worksize.x + LOCAL_SIZE_X-1)/LOCAL_SIZE_X,
+                             (worksize.y + LOCAL_SIZE_Y-1)/LOCAL_SIZE_Y);
+        int totalTiles = ntiles.x*ntiles.y;
+
+        for( tileIdx = groupIdx; tileIdx < totalTiles; tileIdx += ngroups )
+        {
+            int ix0 = (tileIdx % ntiles.x)*LOCAL_SIZE_X;
+            int iy0 = (tileIdx / ntiles.x)*LOCAL_SIZE_Y;
+            int ix = lx, iy = ly;
+            __global const int* psum0 = sum + mad24(iy0, sumstep, ix0) + s->layer_ofs;
+            __global const int* psum1 = psum0 + mad24(iy, sumstep, ix);
+
+            if( ix0 >= worksize.x || iy0 >= worksize.y )
+                continue;
+            #if SUM_BUF_SIZE > 0
+            for( i = lidx*4; i < SUM_BUF_SIZE; i += LOCAL_SIZE_X*LOCAL_SIZE_Y*4 )
+            {
+                int dy = i/SUM_BUF_STEP, dx = i - dy*SUM_BUF_STEP;
+                vstore4(vload4(0, psum0 + mad24(dy, sumstep, dx)), 0, ibuf+i);
+            }
+            #endif
+
+            if( lidx == 0 )
+                lcount[0] = 0;
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            if( ix0 + ix < worksize.x && iy0 + iy < worksize.y )
+            {
+                #if NODE_COUNT==1
+                __global const Stump* stump = (__global const Stump*)nodes;
+                #else
+                __global const Node* node = nodes;
+                __global const float* leaves = leaves0;
+                #endif
+                #if SUM_BUF_SIZE > 0
+                __local const int* psum = ibuf + mad24(iy, SUM_BUF_STEP, ix);
+                #else
+                __global const int* psum = psum1;
+                #endif
+
+                __global const int* psqsum = (__global const int*)(psum1 + sqofs);
+                float sval = (psum[nofs.x] - psum[nofs.y] - psum[nofs.z] + psum[nofs.w])*invarea;
+                float sqval = (psqsum[nofs0.x] - psqsum[nofs0.y] - psqsum[nofs0.z] + psqsum[nofs0.w])*invarea;
+                float nf = (float)normarea * sqrt(max(sqval - sval * sval, 0.f));
+                nf = nf > 0 ? nf : 1.f;
+
+                for( stageIdx = 0; stageIdx < SPLIT_STAGE; stageIdx++ )
+                {
+                    int ntrees = stages[stageIdx].ntrees;
+                    float s = 0.f;
+                    #if NODE_COUNT==1
+                    for( i = 0; i < ntrees; i++ )
+                    {
+                        float4 st = stump[i].st;
+                        __global const OptHaarFeature* f = optfeatures + as_int(st.x);
+                        float4 weight = f->weight;
+
+                        int4 ofs = f->ofs[0];
+                        sval = (psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w])*weight.x;
+                        ofs = f->ofs[1];
+                        sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.y, sval);
+                        if( weight.z > 0 )
+                        {
+                            ofs = f->ofs[2];
+                            sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.z, sval);
+                        }
+
+                        s += (sval < st.y*nf) ? st.z : st.w;
+                    }
+                    stump += ntrees;
+                    #else
+                    for( i = 0; i < ntrees; i++, node += NODE_COUNT, leaves += NODE_COUNT+1 )
+                    {
+                        int idx = 0;
+                        do
+                        {
+                            int4 n = node[idx].n;
+                            __global const OptHaarFeature* f = optfeatures + n.x;
+                            float4 weight = f->weight;
+
+                            int4 ofs = f->ofs[0];
+
+                            sval = (psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w])*weight.x;
+                            ofs = f->ofs[1];
+                            sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.y, sval);
+                            if( weight.z > 0 )
+                            {
+                                ofs = f->ofs[2];
+                                sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.z, sval);
+                            }
+
+                            idx = (sval < as_float(n.y)*nf) ? n.z : n.w;
+                        }
+                        while(idx > 0);
+                        s += leaves[-idx];
+                    }
+                    #endif
+
+                    if( s < stages[stageIdx].threshold )
+                        break;
+                }
+
+                if( stageIdx == SPLIT_STAGE && (ystep == 1 || ((ix | iy) & 1) == 0) )
+                {
+                    int count = atomic_inc(lcount);
+                    lbuf[count] = (int)(ix | (iy << 8));
+                    lnf[count] = nf;
+                }
+            }
+
+            for( stageIdx = SPLIT_STAGE; stageIdx < N_STAGES; stageIdx++ )
+            {
+                barrier(CLK_LOCAL_MEM_FENCE);
+                int nrects = lcount[0];
+
+                if( nrects == 0 )
+                    break;
+                barrier(CLK_LOCAL_MEM_FENCE);
+                if( lidx == 0 )
+                    lcount[0] = 0;
+
+                {
+                    #if NODE_COUNT == 1
+                    __global const Stump* stump = (__global const Stump*)nodes + stages[stageIdx].first;
+                    #else
+                    __global const Node* node = nodes + stages[stageIdx].first*NODE_COUNT;
+                    __global const float* leaves = leaves0 + stages[stageIdx].first*(NODE_COUNT+1);
+                    #endif
+                    int nparts = LOCAL_SIZE / nrects;
+                    int ntrees = stages[stageIdx].ntrees;
+                    int ntrees_p = (ntrees + nparts - 1)/nparts;
+                    int nr = lidx / nparts;
+                    int partidx = -1, idxval = 0;
+                    float partsum = 0.f, nf = 0.f;
+
+                    if( nr < nrects )
+                    {
+                        partidx = lidx % nparts;
+                        idxval = lbuf[nr];
+                        nf = lnf[nr];
+
+                        {
+                        int ntrees0 = ntrees_p*partidx;
+                        int ntrees1 = min(ntrees0 + ntrees_p, ntrees);
+                        int ix1 = idxval & 255, iy1 = idxval >> 8;
+                        #if SUM_BUF_SIZE > 0
+                        __local const int* psum = ibuf + mad24(iy1, SUM_BUF_STEP, ix1);
+                        #else
+                        __global const int* psum = psum0 + mad24(iy1, sumstep, ix1);
+                        #endif
+
+                        #if NODE_COUNT == 1
+                        for( i = ntrees0; i < ntrees1; i++ )
+                        {
+                            float4 st = stump[i].st;
+                            __global const OptHaarFeature* f = optfeatures + as_int(st.x);
+                            float4 weight = f->weight;
+
+                            int4 ofs = f->ofs[0];
+                            float sval = (psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w])*weight.x;
+                            ofs = f->ofs[1];
+                            sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.y, sval);
+                            //if( weight.z > 0 )
+                            if( fabs(weight.z) > 0 )
+                            {
+                                ofs = f->ofs[2];
+                                sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.z, sval);
+                            }
+
+                            partsum += (sval < st.y*nf) ? st.z : st.w;
+                        }
+                        #else
+                        for( i = ntrees0; i < ntrees1; i++ )
+                        {
+                            int idx = 0;
+                            do
+                            {
+                                int4 n = node[i*2 + idx].n;
+                                __global const OptHaarFeature* f = optfeatures + n.x;
+                                float4 weight = f->weight;
+                                int4 ofs = f->ofs[0];
+
+                                float sval = (psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w])*weight.x;
+                                ofs = f->ofs[1];
+                                sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.y, sval);
+                                if( weight.z > 0 )
+                                {
+                                    ofs = f->ofs[2];
+                                    sval = mad((float)(psum[ofs.x] - psum[ofs.y] - psum[ofs.z] + psum[ofs.w]), weight.z, sval);
+                                }
+
+                                idx = (sval < as_float(n.y)*nf) ? n.z : n.w;
+                            }
+                            while(idx > 0);
+                            partsum += leaves[i*3-idx];
+                        }
+                        #endif
+                        }
+                    }
+                    lpartsum[lidx] = partsum;
+                    barrier(CLK_LOCAL_MEM_FENCE);
+
+                    if( partidx == 0 )
+                    {
+                        float s = lpartsum[nr*nparts];
+                        for( i = 1; i < nparts; i++ )
+                            s += lpartsum[i + nr*nparts];
+                        if( s >= stages[stageIdx].threshold )
+                        {
+                            int count = atomic_inc(lcount);
+                            lbuf[count] = idxval;
+                            lnf[count] = nf;
+                        }
+                    }
+                }
+            }
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+            if( stageIdx == N_STAGES )
+            {
+                int nrects = lcount[0];
+                if( lidx < nrects )
+                {
+                    int nfaces = atomic_inc(facepos);
+                    if( nfaces < MAX_FACES )
+                    {
+                        volatile __global int* face = facepos + 1 + nfaces*3;
+                        int val = lbuf[lidx];
+                        face[0] = scaleIdx;
+                        face[1] = ix0 + (val & 255);
+                        face[2] = iy0 + (val >> 8);
+                    }
+                }
+            }
+        }
+    }
+}
+#endif
+
+#ifdef LBP
+#undef CALC_SUM_OFS_
+#define CALC_SUM_OFS_(p0, p1, p2, p3, ptr) \
+    ((ptr)[p0] - (ptr)[p1] - (ptr)[p2] + (ptr)[p3])
+
+__kernel void runLBPClassifierStumpSimple(
+    int nscales, __global const ScaleData* scaleData,
+    __global const int* sum,
+    int _sumstep, int sumoffset,
+    __global const OptLBPFeature* optfeatures,
+    __global const Stage* stages,
+    __global const Stump* stumps,
+    __global const int* bitsets,
+    int bitsetSize,
+
+    volatile __global int* facepos,
+    int2 windowsize)
+{
+    int lx = get_local_id(0);
+    int ly = get_local_id(1);
+    int local_size_x = get_local_size(0);
+    int local_size_y = get_local_size(1);
+    int groupIdx = get_group_id(1)*get_num_groups(0) + get_group_id(0);
+    int ngroups = get_num_groups(0)*get_num_groups(1);
+    int scaleIdx, tileIdx, stageIdx;
+    int sumstep = (int)(_sumstep/sizeof(int));
+
+    for( scaleIdx = nscales-1; scaleIdx >= 0; scaleIdx-- )
+    {
+        __global const ScaleData* s = scaleData + scaleIdx;
+        int ystep = s->ystep;
+        int2 worksize = (int2)(max(s->szi_width - windowsize.x, 0), max(s->szi_height - windowsize.y, 0));
+        int2 ntiles = (int2)((worksize.x/ystep + local_size_x-1)/local_size_x,
+                             (worksize.y/ystep + local_size_y-1)/local_size_y);
+        int totalTiles = ntiles.x*ntiles.y;
+
+        for( tileIdx = groupIdx; tileIdx < totalTiles; tileIdx += ngroups )
+        {
+            int iy = mad24((tileIdx / ntiles.x), local_size_y, ly) * ystep;
+            int ix = mad24((tileIdx % ntiles.x), local_size_x, lx) * ystep;
+
+            if( ix < worksize.x && iy < worksize.y )
+            {
+                __global const int* p = sum + mad24(iy, sumstep, ix) + s->layer_ofs;
+                __global const Stump* stump = stumps;
+                __global const int* bitset = bitsets;
+
+                for( stageIdx = 0; stageIdx < N_STAGES; stageIdx++ )
+                {
+                    int i, ntrees = stages[stageIdx].ntrees;
+                    float s = 0.f;
+                    for( i = 0; i < ntrees; i++, stump++, bitset += bitsetSize )
+                    {
+                        float4 st = stump->st;
+                        __global const OptLBPFeature* f = optfeatures + as_int(st.x);
+                        int16 ofs = f->ofs;
+
+                        int cval = CALC_SUM_OFS_( ofs.s5, ofs.s6, ofs.s9, ofs.sa, p );
+
+                        int mask, idx = (CALC_SUM_OFS_( ofs.s0, ofs.s1, ofs.s4, ofs.s5, p ) >= cval ? 4 : 0); // 0
+                        idx |= (CALC_SUM_OFS_( ofs.s1, ofs.s2, ofs.s5, ofs.s6, p ) >= cval ? 2 : 0); // 1
+                        idx |= (CALC_SUM_OFS_( ofs.s2, ofs.s3, ofs.s6, ofs.s7, p ) >= cval ? 1 : 0); // 2
+
+                        mask = (CALC_SUM_OFS_( ofs.s6, ofs.s7, ofs.sa, ofs.sb, p ) >= cval ? 16 : 0); // 5
+                        mask |= (CALC_SUM_OFS_( ofs.sa, ofs.sb, ofs.se, ofs.sf, p ) >= cval ? 8 : 0);  // 8
+                        mask |= (CALC_SUM_OFS_( ofs.s9, ofs.sa, ofs.sd, ofs.se, p ) >= cval ? 4 : 0);  // 7
+                        mask |= (CALC_SUM_OFS_( ofs.s8, ofs.s9, ofs.sc, ofs.sd, p ) >= cval ? 2 : 0);  // 6
+                        mask |= (CALC_SUM_OFS_( ofs.s4, ofs.s5, ofs.s8, ofs.s9, p ) >= cval ? 1 : 0);  // 7
+
+                        s += (bitset[idx] & (1 << mask)) ? st.z : st.w;
+                    }
+
+                    if( s < stages[stageIdx].threshold )
+                        break;
+                }
+
+                if( stageIdx == N_STAGES )
+                {
+                    int nfaces = atomic_inc(facepos);
+                    if( nfaces < MAX_FACES )
+                    {
+                        volatile __global int* face = facepos + 1 + nfaces*3;
+                        face[0] = scaleIdx;
+                        face[1] = ix;
+                        face[2] = iy;
+                    }
+                }
+            }
+        }
+    }
+}
+
+__kernel __attribute__((reqd_work_group_size(LOCAL_SIZE_X,LOCAL_SIZE_Y,1)))
+void runLBPClassifierStump(
+    int nscales, __global const ScaleData* scaleData,
+    __global const int* sum,
+    int _sumstep, int sumoffset,
+    __global const OptLBPFeature* optfeatures,
+    __global const Stage* stages,
+    __global const Stump* stumps,
+    __global const int* bitsets,
+    int bitsetSize,
+
+    volatile __global int* facepos,
+    int2 windowsize)
+{
+    int lx = get_local_id(0);
+    int ly = get_local_id(1);
+    int groupIdx = get_group_id(0);
+    int i, ngroups = get_global_size(0)/LOCAL_SIZE_X;
+    int scaleIdx, tileIdx, stageIdx;
+    int sumstep = (int)(_sumstep/sizeof(int));
+    int lidx = ly*LOCAL_SIZE_X + lx;
+
+    #define LOCAL_SIZE (LOCAL_SIZE_X*LOCAL_SIZE_Y)
+    __local int lstore[SUM_BUF_SIZE + LOCAL_SIZE*3/2+1];
+    #if SUM_BUF_SIZE > 0
+    __local int* ibuf = lstore;
+    __local int* lcount = ibuf + SUM_BUF_SIZE;
+    #else
+    __local int* lcount = lstore;
+    #endif
+    __local float* lpartsum = (__local float*)(lcount + 1);
+    __local short* lbuf = (__local short*)(lpartsum + LOCAL_SIZE);
+
+    for( scaleIdx = nscales-1; scaleIdx >= 0; scaleIdx-- )
+    {
+        __global const ScaleData* s = scaleData + scaleIdx;
+        int ystep = s->ystep;
+        int2 worksize = (int2)(max(s->szi_width - windowsize.x, 0), max(s->szi_height - windowsize.y, 0));
+        int2 ntiles = (int2)((worksize.x + LOCAL_SIZE_X-1)/LOCAL_SIZE_X,
+                             (worksize.y + LOCAL_SIZE_Y-1)/LOCAL_SIZE_Y);
+        int totalTiles = ntiles.x*ntiles.y;
+
+        for( tileIdx = groupIdx; tileIdx < totalTiles; tileIdx += ngroups )
+        {
+            int ix0 = (tileIdx % ntiles.x)*LOCAL_SIZE_X;
+            int iy0 = (tileIdx / ntiles.x)*LOCAL_SIZE_Y;
+            int ix = lx, iy = ly;
+            __global const int* psum0 = sum + mad24(iy0, sumstep, ix0) + s->layer_ofs;
+
+            if( ix0 >= worksize.x || iy0 >= worksize.y )
+                continue;
+            #if SUM_BUF_SIZE > 0
+            for( i = lidx*4; i < SUM_BUF_SIZE; i += LOCAL_SIZE_X*LOCAL_SIZE_Y*4 )
+            {
+                int dy = i/SUM_BUF_STEP, dx = i - dy*SUM_BUF_STEP;
+                vstore4(vload4(0, psum0 + mad24(dy, sumstep, dx)), 0, ibuf+i);
+            }
+            barrier(CLK_LOCAL_MEM_FENCE);
+            #endif
+
+            if( lidx == 0 )
+                lcount[0] = 0;
+            barrier(CLK_LOCAL_MEM_FENCE);
+
+            if( ix0 + ix < worksize.x && iy0 + iy < worksize.y )
+            {
+                __global const Stump* stump = stumps;
+                __global const int* bitset = bitsets;
+                #if SUM_BUF_SIZE > 0
+                __local const int* p = ibuf + mad24(iy, SUM_BUF_STEP, ix);
+                #else
+                __global const int* p = psum0 + mad24(iy, sumstep, ix);
+                #endif
+
+                for( stageIdx = 0; stageIdx < SPLIT_STAGE; stageIdx++ )
+                {
+                    int ntrees = stages[stageIdx].ntrees;
+                    float s = 0.f;
+                    for( i = 0; i < ntrees; i++, stump++, bitset += bitsetSize )
+                    {
+                        float4 st = stump->st;
+                        __global const OptLBPFeature* f = optfeatures + as_int(st.x);
+                        int16 ofs = f->ofs;
+
+                        int cval = CALC_SUM_OFS_( ofs.s5, ofs.s6, ofs.s9, ofs.sa, p );
+
+                        int mask, idx = (CALC_SUM_OFS_( ofs.s0, ofs.s1, ofs.s4, ofs.s5, p ) >= cval ? 4 : 0); // 0
+                        idx |= (CALC_SUM_OFS_( ofs.s1, ofs.s2, ofs.s5, ofs.s6, p ) >= cval ? 2 : 0); // 1
+                        idx |= (CALC_SUM_OFS_( ofs.s2, ofs.s3, ofs.s6, ofs.s7, p ) >= cval ? 1 : 0); // 2
+
+                        mask = (CALC_SUM_OFS_( ofs.s6, ofs.s7, ofs.sa, ofs.sb, p ) >= cval ? 16 : 0); // 5
+                        mask |= (CALC_SUM_OFS_( ofs.sa, ofs.sb, ofs.se, ofs.sf, p ) >= cval ? 8 : 0);  // 8
+                        mask |= (CALC_SUM_OFS_( ofs.s9, ofs.sa, ofs.sd, ofs.se, p ) >= cval ? 4 : 0);  // 7
+                        mask |= (CALC_SUM_OFS_( ofs.s8, ofs.s9, ofs.sc, ofs.sd, p ) >= cval ? 2 : 0);  // 6
+                        mask |= (CALC_SUM_OFS_( ofs.s4, ofs.s5, ofs.s8, ofs.s9, p ) >= cval ? 1 : 0);  // 7
+
+                        s += (bitset[idx] & (1 << mask)) ? st.z : st.w;
+                    }
+
+                    if( s < stages[stageIdx].threshold )
+                        break;
+                }
+
+                if( stageIdx == SPLIT_STAGE && (ystep == 1 || ((ix | iy) & 1) == 0) )
+                {
+                    int count = atomic_inc(lcount);
+                    lbuf[count] = (int)(ix | (iy << 8));
+                }
+            }
+
+            for( stageIdx = SPLIT_STAGE; stageIdx < N_STAGES; stageIdx++ )
+            {
+                int nrects = lcount[0];
+
+                barrier(CLK_LOCAL_MEM_FENCE);
+                if( nrects == 0 )
+                    break;
+                if( lidx == 0 )
+                    lcount[0] = 0;
+
+                {
+                    __global const Stump* stump = stumps + stages[stageIdx].first;
+                    __global const int* bitset = bitsets + stages[stageIdx].first*bitsetSize;
+                    int nparts = LOCAL_SIZE / nrects;
+                    int ntrees = stages[stageIdx].ntrees;
+                    int ntrees_p = (ntrees + nparts - 1)/nparts;
+                    int nr = lidx / nparts;
+                    int partidx = -1, idxval = 0;
+                    float partsum = 0.f, nf = 0.f;
+
+                    if( nr < nrects )
+                    {
+                        partidx = lidx % nparts;
+                        idxval = lbuf[nr];
+
+                        {
+                            int ntrees0 = ntrees_p*partidx;
+                            int ntrees1 = min(ntrees0 + ntrees_p, ntrees);
+                            int ix1 = idxval & 255, iy1 = idxval >> 8;
+                            #if SUM_BUF_SIZE > 0
+                            __local const int* p = ibuf + mad24(iy1, SUM_BUF_STEP, ix1);
+                            #else
+                            __global const int* p = psum0 + mad24(iy1, sumstep, ix1);
+                            #endif
+
+                            for( i = ntrees0; i < ntrees1; i++ )
+                            {
+                                float4 st = stump[i].st;
+                                __global const OptLBPFeature* f = optfeatures + as_int(st.x);
+                                int16 ofs = f->ofs;
+
+                                #define CALC_SUM_OFS_(p0, p1, p2, p3, ptr) \
+                                    ((ptr)[p0] - (ptr)[p1] - (ptr)[p2] + (ptr)[p3])
+
+                                int cval = CALC_SUM_OFS_( ofs.s5, ofs.s6, ofs.s9, ofs.sa, p );
+
+                                int mask, idx = (CALC_SUM_OFS_( ofs.s0, ofs.s1, ofs.s4, ofs.s5, p ) >= cval ? 4 : 0); // 0
+                                idx |= (CALC_SUM_OFS_( ofs.s1, ofs.s2, ofs.s5, ofs.s6, p ) >= cval ? 2 : 0); // 1
+                                idx |= (CALC_SUM_OFS_( ofs.s2, ofs.s3, ofs.s6, ofs.s7, p ) >= cval ? 1 : 0); // 2
+
+                                mask = (CALC_SUM_OFS_( ofs.s6, ofs.s7, ofs.sa, ofs.sb, p ) >= cval ? 16 : 0); // 5
+                                mask |= (CALC_SUM_OFS_( ofs.sa, ofs.sb, ofs.se, ofs.sf, p ) >= cval ? 8 : 0);  // 8
+                                mask |= (CALC_SUM_OFS_( ofs.s9, ofs.sa, ofs.sd, ofs.se, p ) >= cval ? 4 : 0);  // 7
+                                mask |= (CALC_SUM_OFS_( ofs.s8, ofs.s9, ofs.sc, ofs.sd, p ) >= cval ? 2 : 0);  // 6
+                                mask |= (CALC_SUM_OFS_( ofs.s4, ofs.s5, ofs.s8, ofs.s9, p ) >= cval ? 1 : 0);  // 7
+
+                                partsum += (bitset[i*bitsetSize + idx] & (1 << mask)) ? st.z : st.w;
+                            }
+                        }
+                    }
+                    lpartsum[lidx] = partsum;
+                    barrier(CLK_LOCAL_MEM_FENCE);
+
+                    if( partidx == 0 )
+                    {
+                        float s = lpartsum[nr*nparts];
+                        for( i = 1; i < nparts; i++ )
+                            s += lpartsum[i + nr*nparts];
+                        if( s >= stages[stageIdx].threshold )
+                        {
+                            int count = atomic_inc(lcount);
+                            lbuf[count] = idxval;
+                        }
+                    }
+                }
+            }
+
+            barrier(CLK_LOCAL_MEM_FENCE);
+            if( stageIdx == N_STAGES )
+            {
+                int nrects = lcount[0];
+                if( lidx < nrects )
+                {
+                    int nfaces = atomic_inc(facepos);
+                    if( nfaces < MAX_FACES )
+                    {
+                        volatile __global int* face = facepos + 1 + nfaces*3;
+                        int val = lbuf[lidx];
+                        face[0] = scaleIdx;
+                        face[1] = ix0 + (val & 255);
+                        face[2] = iy0 + (val >> 8);
+                    }
+                }
+            }
+        }
+    }
+}
+#endif

modules/xobjdetect/src/opencl/objdetect_hog.cl

@@ -0,0 +1,633 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Wenju He, wenju@multicorewareinc.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#define CELL_WIDTH 8
+#define CELL_HEIGHT 8
+#define CELLS_PER_BLOCK_X 2
+#define CELLS_PER_BLOCK_Y 2
+#define NTHREADS 256
+#define CV_PI_F M_PI_F
+
+#ifdef INTEL_DEVICE
+#define QANGLE_TYPE     int
+#define QANGLE_TYPE2    int2
+#else
+#define QANGLE_TYPE     uchar
+#define QANGLE_TYPE2    uchar2
+#endif
+
+//----------------------------------------------------------------------------
+// Histogram computation
+// 12 threads for a cell, 12x4 threads per block
+// Use pre-computed gaussian and interp_weight lookup tables
+__kernel void compute_hists_lut_kernel(
+    const int cblock_stride_x, const int cblock_stride_y,
+    const int cnbins, const int cblock_hist_size, const int img_block_width,
+    const int blocks_in_group, const int blocks_total,
+    const int grad_quadstep, const int qangle_step,
+    __global const float* grad, __global const QANGLE_TYPE* qangle,
+    __global const float* gauss_w_lut,
+    __global float* block_hists, __local float* smem)
+{
+    const int lx = get_local_id(0);
+    const int lp = lx / 24; /* local group id */
+    const int gid = get_group_id(0) * blocks_in_group + lp;/* global group id */
+    const int gidY = gid / img_block_width;
+    const int gidX = gid - gidY * img_block_width;
+
+    const int lidX = lx - lp * 24;
+    const int lidY = get_local_id(1);
+
+    const int cell_x = lidX / 12;
+    const int cell_y = lidY;
+    const int cell_thread_x = lidX - cell_x * 12;
+
+    __local float* hists = smem + lp * cnbins * (CELLS_PER_BLOCK_X *
+        CELLS_PER_BLOCK_Y * 12 + CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y);
+    __local float* final_hist = hists + cnbins *
+        (CELLS_PER_BLOCK_X * CELLS_PER_BLOCK_Y * 12);
+
+    const int offset_x = gidX * cblock_stride_x + (cell_x << 2) + cell_thread_x;
+    const int offset_y = gidY * cblock_stride_y + (cell_y << 2);
+
+    __global const float* grad_ptr = (gid < blocks_total) ?
+        grad + offset_y * grad_quadstep + (offset_x << 1) : grad;
+    __global const QANGLE_TYPE* qangle_ptr = (gid < blocks_total) ?
+        qangle + offset_y * qangle_step + (offset_x << 1) : qangle;
+
+    __local float* hist = hists + 12 * (cell_y * CELLS_PER_BLOCK_Y + cell_x) +
+        cell_thread_x;
+    for (int bin_id = 0; bin_id < cnbins; ++bin_id)
+        hist[bin_id * 48] = 0.f;
+
+    const int dist_x = -4 + cell_thread_x - 4 * cell_x;
+    const int dist_center_x = dist_x - 4 * (1 - 2 * cell_x);
+
+    const int dist_y_begin = -4 - 4 * lidY;
+    for (int dist_y = dist_y_begin; dist_y < dist_y_begin + 12; ++dist_y)
+    {
+        float2 vote = (float2) (grad_ptr[0], grad_ptr[1]);
+        QANGLE_TYPE2 bin = (QANGLE_TYPE2) (qangle_ptr[0], qangle_ptr[1]);
+
+        grad_ptr += grad_quadstep;
+        qangle_ptr += qangle_step;
+
+        int dist_center_y = dist_y - 4 * (1 - 2 * cell_y);
+
+        int idx = (dist_center_y + 8) * 16 + (dist_center_x + 8);
+        float gaussian = gauss_w_lut[idx];
+        idx = (dist_y + 8) * 16 + (dist_x + 8);
+        float interp_weight = gauss_w_lut[256+idx];
+
+        hist[bin.x * 48] += gaussian * interp_weight * vote.x;
+        hist[bin.y * 48] += gaussian * interp_weight * vote.y;
+    }
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    volatile __local float* hist_ = hist;
+    for (int bin_id = 0; bin_id < cnbins; ++bin_id, hist_ += 48)
+    {
+        if (cell_thread_x < 6)
+            hist_[0] += hist_[6];
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if (cell_thread_x < 3)
+            hist_[0] += hist_[3];
+        barrier(CLK_LOCAL_MEM_FENCE);
+        if (cell_thread_x == 0)
+            final_hist[(cell_x * 2 + cell_y) * cnbins + bin_id] =
+                hist_[0] + hist_[1] + hist_[2];
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    int tid = (cell_y * CELLS_PER_BLOCK_Y + cell_x) * 12 + cell_thread_x;
+    if ((tid < cblock_hist_size) && (gid < blocks_total))
+    {
+        __global float* block_hist = block_hists +
+            (gidY * img_block_width + gidX) * cblock_hist_size;
+        block_hist[tid] = final_hist[tid];
+    }
+}
+
+//-------------------------------------------------------------
+//  Normalization of histograms via L2Hys_norm
+//  optimized for the case of 9 bins
+__kernel void normalize_hists_36_kernel(__global float* block_hists,
+                                        const float threshold, __local float *squares)
+{
+    const int tid = get_local_id(0);
+    const int gid = get_global_id(0);
+    const int bid = tid / 36;      /* block-hist id, (0 - 6) */
+    const int boffset = bid * 36;  /* block-hist offset in the work-group */
+    const int hid = tid - boffset; /* histogram bin id, (0 - 35) */
+
+    float elem = block_hists[gid];
+    squares[tid] = elem * elem;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    __local float* smem = squares + boffset;
+    float sum = smem[hid];
+    if (hid < 18)
+        smem[hid] = sum = sum + smem[hid + 18];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 9)
+        smem[hid] = sum = sum + smem[hid + 9];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 4)
+        smem[hid] = sum + smem[hid + 4];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    sum = smem[0] + smem[1] + smem[2] + smem[3] + smem[8];
+
+    elem = elem / (sqrt(sum) + 3.6f);
+    elem = min(elem, threshold);
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    squares[tid] = elem * elem;
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    sum = smem[hid];
+    if (hid < 18)
+      smem[hid] = sum = sum + smem[hid + 18];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 9)
+        smem[hid] = sum = sum + smem[hid + 9];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (hid < 4)
+        smem[hid] = sum + smem[hid + 4];
+    barrier(CLK_LOCAL_MEM_FENCE);
+    sum = smem[0] + smem[1] + smem[2] + smem[3] + smem[8];
+
+    block_hists[gid] = elem / (sqrt(sum) + 1e-3f);
+}
+
+//-------------------------------------------------------------
+//  Normalization of histograms via L2Hys_norm
+//
+inline float reduce_smem(volatile __local float* smem, int size)
+{
+    unsigned int tid = get_local_id(0);
+    float sum = smem[tid];
+
+    if (size >= 512) { if (tid < 256) smem[tid] = sum = sum + smem[tid + 256];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 256) { if (tid < 128) smem[tid] = sum = sum + smem[tid + 128];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 128) { if (tid < 64) smem[tid] = sum = sum + smem[tid + 64];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 64) { if (tid < 32) smem[tid] = sum = sum + smem[tid + 32];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 32) { if (tid < 16) smem[tid] = sum = sum + smem[tid + 16];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 16) { if (tid < 8) smem[tid] = sum = sum + smem[tid + 8];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 8) { if (tid < 4) smem[tid] = sum = sum + smem[tid + 4];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 4) { if (tid < 2) smem[tid] = sum = sum + smem[tid + 2];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+    if (size >= 2) { if (tid < 1) smem[tid] = sum = sum + smem[tid + 1];
+        barrier(CLK_LOCAL_MEM_FENCE); }
+
+    return sum;
+}
+
+__kernel void normalize_hists_kernel(
+    const int nthreads, const int block_hist_size, const int img_block_width,
+    __global float* block_hists, const float threshold, __local float *squares)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global float* hist = block_hists + (gidY * img_block_width + gidX) *
+        block_hist_size + tid;
+
+    float elem = 0.f;
+    if (tid < block_hist_size)
+        elem = hist[0];
+
+    squares[tid] = elem * elem;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    float sum = reduce_smem(squares, nthreads);
+
+    float scale = 1.0f / (sqrt(sum) + 0.1f * block_hist_size);
+    elem = min(elem * scale, threshold);
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    squares[tid] = elem * elem;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    sum = reduce_smem(squares, nthreads);
+    scale = 1.0f / (sqrt(sum) + 1e-3f);
+
+    if (tid < block_hist_size)
+        hist[0] = elem * scale;
+}
+
+#define reduce_with_sync(target, sharedMemory, localMemory, tid, offset) \
+    if (tid < target) sharedMemory[tid] = localMemory = localMemory + sharedMemory[tid + offset]; \
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+//---------------------------------------------------------------------
+//  Linear SVM based classification
+//  48x96 window, 9 bins and default parameters
+//  180 threads, each thread corresponds to a bin in a row
+__kernel void classify_hists_180_kernel(
+    const int cdescr_width, const int cdescr_height, const int cblock_hist_size,
+    const int img_win_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float * block_hists, __global const float* coefs,
+    float free_coef, float threshold, __global uchar* labels)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    float product = 0.f;
+
+    for (int i = 0; i < cdescr_height; i++)
+    {
+        product += coefs[i * cdescr_width + tid] *
+            hist[i * img_block_width * cblock_hist_size + tid];
+    }
+
+    __local float products[180];
+
+    products[tid] = product;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    reduce_with_sync(90, products, product, tid, 90);
+    reduce_with_sync(45, products, product, tid, 45);
+    reduce_with_sync(13, products, product, tid, 32); // 13 is not typo
+    reduce_with_sync(16, products, product, tid, 16);
+    reduce_with_sync(8, products, product, tid, 8);
+    reduce_with_sync(4, products, product, tid, 4);
+    reduce_with_sync(2, products, product, tid, 2);
+
+    if (tid == 0){
+        product = product + products[tid + 1];
+        labels[gidY * img_win_width + gidX] = (product + free_coef >= threshold);
+    }
+}
+
+//---------------------------------------------------------------------
+//  Linear SVM based classification
+//  64x128 window, 9 bins and default parameters
+//  256 threads, 252 of them are used
+__kernel void classify_hists_252_kernel(
+    const int cdescr_width, const int cdescr_height, const int cblock_hist_size,
+    const int img_win_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float * block_hists, __global const float* coefs,
+    float free_coef, float threshold, __global uchar* labels)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    float product = 0.f;
+    if (tid < cdescr_width)
+    {
+        for (int i = 0; i < cdescr_height; i++)
+            product += coefs[i * cdescr_width + tid] *
+                hist[i * img_block_width * cblock_hist_size + tid];
+    }
+
+    __local float products[NTHREADS];
+
+    products[tid] = product;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    reduce_with_sync(128, products, product, tid, 128);
+    reduce_with_sync(64, products, product, tid, 64);
+    reduce_with_sync(32, products, product, tid, 32);
+    reduce_with_sync(16, products, product, tid, 16);
+    reduce_with_sync(8, products, product, tid, 8);
+    reduce_with_sync(4, products, product, tid, 4);
+    reduce_with_sync(2, products, product, tid, 2);
+
+    if (tid == 0){
+        product = product + products[tid + 1];
+        labels[gidY * img_win_width + gidX] = (product + free_coef >= threshold);
+    }
+}
+
+//---------------------------------------------------------------------
+//  Linear SVM based classification
+//  256 threads
+__kernel void classify_hists_kernel(
+    const int cdescr_size, const int cdescr_width, const int cblock_hist_size,
+    const int img_win_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float * block_hists, __global const float* coefs,
+    float free_coef, float threshold, __global uchar* labels)
+{
+    const int tid = get_local_id(0);
+    const int gidX = get_group_id(0);
+    const int gidY = get_group_id(1);
+
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    float product = 0.f;
+    for (int i = tid; i < cdescr_size; i += NTHREADS)
+    {
+        int offset_y = i / cdescr_width;
+        int offset_x = i - offset_y * cdescr_width;
+        product += coefs[i] *
+            hist[offset_y * img_block_width * cblock_hist_size + offset_x];
+    }
+
+    __local float products[NTHREADS];
+
+    products[tid] = product;
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    reduce_with_sync(128, products, product, tid, 128);
+    reduce_with_sync(64, products, product, tid, 64);
+    reduce_with_sync(32, products, product, tid, 32);
+    reduce_with_sync(16, products, product, tid, 16);
+    reduce_with_sync(8, products, product, tid, 8);
+    reduce_with_sync(4, products, product, tid, 4);
+    reduce_with_sync(2, products, product, tid, 2);
+
+    if (tid == 0){
+        products[tid] = product = product + products[tid + 1];
+        labels[gidY * img_win_width + gidX] = (product + free_coef >= threshold);
+    }
+}
+
+//----------------------------------------------------------------------------
+// Extract descriptors
+
+__kernel void extract_descrs_by_rows_kernel(
+    const int cblock_hist_size, const int descriptors_quadstep,
+    const int cdescr_size, const int cdescr_width, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float* block_hists, __global float* descriptors)
+{
+    int tid = get_local_id(0);
+    int gidX = get_group_id(0);
+    int gidY = get_group_id(1);
+
+    // Get left top corner of the window in src
+    __global const float* hist = block_hists + (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    // Get left top corner of the window in dst
+    __global float* descriptor = descriptors +
+        (gidY * get_num_groups(0) + gidX) * descriptors_quadstep;
+
+    // Copy elements from src to dst
+    for (int i = tid; i < cdescr_size; i += NTHREADS)
+    {
+        int offset_y = i / cdescr_width;
+        int offset_x = i - offset_y * cdescr_width;
+        descriptor[i] = hist[offset_y * img_block_width * cblock_hist_size + offset_x];
+    }
+}
+
+__kernel void extract_descrs_by_cols_kernel(
+    const int cblock_hist_size, const int descriptors_quadstep, const int cdescr_size,
+    const int cnblocks_win_x, const int cnblocks_win_y, const int img_block_width,
+    const int win_block_stride_x, const int win_block_stride_y,
+    __global const float* block_hists, __global float* descriptors)
+{
+    int tid = get_local_id(0);
+    int gidX = get_group_id(0);
+    int gidY = get_group_id(1);
+
+    // Get left top corner of the window in src
+    __global const float* hist = block_hists +  (gidY * win_block_stride_y *
+        img_block_width + gidX * win_block_stride_x) * cblock_hist_size;
+
+    // Get left top corner of the window in dst
+    __global float* descriptor = descriptors +
+        (gidY * get_num_groups(0) + gidX) * descriptors_quadstep;
+
+    // Copy elements from src to dst
+    for (int i = tid; i < cdescr_size; i += NTHREADS)
+    {
+        int block_idx = i / cblock_hist_size;
+        int idx_in_block = i - block_idx * cblock_hist_size;
+
+        int y = block_idx / cnblocks_win_x;
+        int x = block_idx - y * cnblocks_win_x;
+
+        descriptor[(x * cnblocks_win_y + y) * cblock_hist_size + idx_in_block] =
+            hist[(y * img_block_width  + x) * cblock_hist_size + idx_in_block];
+    }
+}
+
+//----------------------------------------------------------------------------
+// Gradients computation
+
+__kernel void compute_gradients_8UC4_kernel(
+    const int height, const int width,
+    const int img_step, const int grad_quadstep, const int qangle_step,
+    const __global uchar4 * img, __global float * grad, __global QANGLE_TYPE * qangle,
+    const float angle_scale, const char correct_gamma, const int cnbins)
+{
+    const int x = get_global_id(0);
+    const int tid = get_local_id(0);
+    const int gSizeX = get_local_size(0);
+    const int gidY = get_group_id(1);
+
+    __global const uchar4* row = img + gidY * img_step;
+
+    __local float sh_row[(NTHREADS + 2) * 3];
+
+    uchar4 val;
+    if (x < width)
+        val = row[x];
+    else
+        val = row[width - 2];
+
+    sh_row[tid + 1] = val.x;
+    sh_row[tid + 1 + (NTHREADS + 2)] = val.y;
+    sh_row[tid + 1 + 2 * (NTHREADS + 2)] = val.z;
+
+    if (tid == 0)
+    {
+        val = row[max(x - 1, 1)];
+        sh_row[0] = val.x;
+        sh_row[(NTHREADS + 2)] = val.y;
+        sh_row[2 * (NTHREADS + 2)] = val.z;
+    }
+
+    if (tid == gSizeX - 1)
+    {
+        val = row[min(x + 1, width - 2)];
+        sh_row[gSizeX + 1] = val.x;
+        sh_row[gSizeX + 1 + (NTHREADS + 2)] = val.y;
+        sh_row[gSizeX + 1 + 2 * (NTHREADS + 2)] = val.z;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (x < width)
+    {
+        float4 a = (float4) (sh_row[tid], sh_row[tid + (NTHREADS + 2)],
+            sh_row[tid + 2 * (NTHREADS + 2)], 0);
+        float4 b = (float4) (sh_row[tid + 2], sh_row[tid + 2 + (NTHREADS + 2)],
+            sh_row[tid + 2 + 2 * (NTHREADS + 2)], 0);
+
+        float4 dx;
+        if (correct_gamma == 1)
+            dx = sqrt(b) - sqrt(a);
+        else
+            dx = b - a;
+
+        float4 dy = (float4) 0.f;
+
+        if (gidY > 0 && gidY < height - 1)
+        {
+            a = convert_float4(img[(gidY - 1) * img_step + x].xyzw);
+            b = convert_float4(img[(gidY + 1) * img_step + x].xyzw);
+
+            if (correct_gamma == 1)
+                dy = sqrt(b) - sqrt(a);
+            else
+                dy = b - a;
+        }
+
+        float4 mag = hypot(dx, dy);
+        float best_dx = dx.x;
+        float best_dy = dy.x;
+
+        float mag0 = mag.x;
+        if (mag0 < mag.y)
+        {
+            best_dx = dx.y;
+            best_dy = dy.y;
+            mag0 = mag.y;
+        }
+
+        if (mag0 < mag.z)
+        {
+            best_dx = dx.z;
+            best_dy = dy.z;
+            mag0 = mag.z;
+        }
+
+        float ang = (atan2(best_dy, best_dx) + CV_PI_F) * angle_scale - 0.5f;
+        int hidx = (int)floor(ang);
+        ang -= hidx;
+        hidx = (hidx + cnbins) % cnbins;
+
+        qangle[(gidY * qangle_step + x) << 1] = hidx;
+        qangle[((gidY * qangle_step + x) << 1) + 1] = (hidx + 1) % cnbins;
+        grad[(gidY * grad_quadstep + x) << 1] = mag0 * (1.f - ang);
+        grad[((gidY * grad_quadstep + x) << 1) + 1] = mag0 * ang;
+    }
+}
+
+__kernel void compute_gradients_8UC1_kernel(
+    const int height, const int width,
+    const int img_step, const int grad_quadstep, const int qangle_step,
+    __global const uchar * img, __global float * grad, __global QANGLE_TYPE * qangle,
+    const float angle_scale, const char correct_gamma, const int cnbins)
+{
+    const int x = get_global_id(0);
+    const int tid = get_local_id(0);
+    const int gSizeX = get_local_size(0);
+    const int gidY = get_group_id(1);
+
+    __global const uchar* row = img + gidY * img_step;
+
+    __local float sh_row[NTHREADS + 2];
+
+    if (x < width)
+        sh_row[tid + 1] = row[x];
+    else
+        sh_row[tid + 1] = row[width - 2];
+
+    if (tid == 0)
+        sh_row[0] = row[max(x - 1, 1)];
+
+    if (tid == gSizeX - 1)
+        sh_row[gSizeX + 1] = row[min(x + 1, width - 2)];
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (x < width)
+    {
+        float dx;
+
+        if (correct_gamma == 1)
+            dx = sqrt(sh_row[tid + 2]) - sqrt(sh_row[tid]);
+        else
+            dx = sh_row[tid + 2] - sh_row[tid];
+
+        float dy = 0.f;
+        if (gidY > 0 && gidY < height - 1)
+        {
+            float a = (float) img[ (gidY + 1) * img_step + x ];
+            float b = (float) img[ (gidY - 1) * img_step + x ];
+            if (correct_gamma == 1)
+                dy = sqrt(a) - sqrt(b);
+            else
+                dy = a - b;
+        }
+        float mag = hypot(dx, dy);
+
+        float ang = (atan2(dy, dx) + CV_PI_F) * angle_scale - 0.5f;
+        int hidx = (int)floor(ang);
+        ang -= hidx;
+        hidx = (hidx + cnbins) % cnbins;
+
+        qangle[ (gidY * qangle_step + x) << 1 ]     = hidx;
+        qangle[ ((gidY * qangle_step + x) << 1) + 1 ] = (hidx + 1) % cnbins;
+        grad[ (gidY * grad_quadstep + x) << 1 ]       = mag * (1.f - ang);
+        grad[ ((gidY * grad_quadstep + x) << 1) + 1 ]   = mag * ang;
+    }
+}

modules/xobjdetect/src/precomp.hpp

@@ -0,0 +1,72 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#ifndef __OPENCV_XOBJDETECT_PRECOMP_HPP__
+#define __OPENCV_XOBJDETECT_PRECOMP_HPP__
+
+#include <opencv2/xobjdetect.hpp>
+#include <opencv2/core/utility.hpp>
+#include <opencv2/core/ocl.hpp>
+#include "opencv2/core/private.hpp"
+#include <opencv2/imgproc.hpp>
+#include <opencv2/core.hpp>
+#include <opencv2/imgcodecs.hpp>
+
+#include <algorithm>
+#include <vector>
+#include <string>
+#include <cmath>
+#include <iostream>
+#include <fstream>
+#include <iomanip>
+#include <sstream>
+#include <cassert>
+#include <cstdio>
+
+#include "feature_evaluator.hpp"
+#include "lbpfeatures.h"
+#include "waldboost.hpp"
+#include "wbdetector.hpp"
+
+#endif /* __OPENCV_XOBJDETECT_PRECOMP_HPP__ */

modules/xobjdetect/src/waldboost.cpp

@@ -0,0 +1,454 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#include "precomp.hpp"
+
+namespace cv {
+namespace xobjdetect {
+
+static void compute_cdf(const Mat1b& data,
+                        const Mat1f& weights,
+                        Mat1f& cdf)
+{
+    for (int i = 0; i < cdf.cols; ++i)
+        cdf(0, i) = 0;
+
+    for (int i = 0; i < weights.cols; ++i) {
+        cdf(0, data(0, i)) += weights(0, i);
+    }
+
+    for (int i = 1; i < cdf.cols; ++i) {
+        cdf(0, i) += cdf(0, i - 1);
+    }
+}
+
+static void compute_min_step(const Mat &data_pos, const Mat &data_neg, size_t n_bins,
+                      Mat &data_min, Mat &data_step)
+{
+    // Check that quantized data will fit in unsigned char
+    assert(n_bins <= 256);
+
+    assert(data_pos.rows == data_neg.rows);
+
+    Mat reduced_pos, reduced_neg;
+
+    reduce(data_pos, reduced_pos, 1, REDUCE_MIN);
+    reduce(data_neg, reduced_neg, 1, REDUCE_MIN);
+    min(reduced_pos, reduced_neg, data_min);
+    data_min -= 0.01;
+
+    Mat data_max;
+    reduce(data_pos, reduced_pos, 1, REDUCE_MAX);
+    reduce(data_neg, reduced_neg, 1, REDUCE_MAX);
+    max(reduced_pos, reduced_neg, data_max);
+    data_max += 0.01;
+
+    data_step = (data_max - data_min) / (double)(n_bins - 1);
+}
+
+static void quantize_data(Mat &data, Mat1f &data_min, Mat1f &data_step)
+{
+//#pragma omp parallel for
+    for (int col = 0; col < data.cols; ++col) {
+        data.col(col) -= data_min;
+        data.col(col) /= data_step;
+    }
+    data.convertTo(data, CV_8U);
+}
+
+WaldBoost::WaldBoost(int weak_count):
+    weak_count_(weak_count),
+    thresholds_(),
+    alphas_(),
+    feature_indices_(),
+    polarities_(),
+    cascade_thresholds_() {}
+
+WaldBoost::WaldBoost():
+    weak_count_(),
+    thresholds_(),
+    alphas_(),
+    feature_indices_(),
+    polarities_(),
+    cascade_thresholds_() {}
+
+std::vector<int> WaldBoost::get_feature_indices()
+{
+    return feature_indices_;
+}
+
+void WaldBoost::detect(Ptr<CvFeatureEvaluator> eval,
+            const Mat& img, const std::vector<float>& scales,
+            std::vector<Rect>& bboxes, Mat1f& confidences)
+{
+    bboxes.clear();
+    confidences.release();
+
+    Mat resized_img;
+    int step = 4;
+    float h;
+    for (size_t i = 0; i < scales.size(); ++i) {
+        float scale = scales[i];
+        resize(img, resized_img, Size(), scale, scale, INTER_LINEAR_EXACT);
+        eval->setImage(resized_img, 0, 0, feature_indices_);
+        int n_rows = (int)(24 / scale);
+        int n_cols = (int)(24 / scale);
+        for (int r = 0; r + 24 < resized_img.rows; r += step) {
+            for (int c = 0; c + 24 < resized_img.cols; c += step) {
+                //eval->setImage(resized_img(Rect(c, r, 24, 24)), 0, 0);
+                eval->setWindow(Point(c, r));
+                if (predict(eval, &h) == +1) {
+                    int row = (int)(r / scale);
+                    int col = (int)(c / scale);
+                    bboxes.push_back(Rect(col, row, n_cols, n_rows));
+                    confidences.push_back(h);
+                }
+            }
+        }
+    }
+    groupRectangles(bboxes, 3, 0.7);
+}
+
+void WaldBoost::detect(Ptr<CvFeatureEvaluator> eval,
+            const Mat& img, const std::vector<float>& scales,
+            std::vector<Rect>& bboxes, std::vector<double>& confidences)
+{
+    bboxes.clear();
+    confidences.clear();
+
+    Mat resized_img;
+    int step = 4;
+    float h;
+    for (size_t i = 0; i < scales.size(); ++i) {
+        float scale = scales[i];
+        resize(img, resized_img, Size(), scale, scale, INTER_LINEAR_EXACT);
+        eval->setImage(resized_img, 0, 0, feature_indices_);
+        int n_rows = (int)(24 / scale);
+        int n_cols = (int)(24 / scale);
+        for (int r = 0; r + 24 < resized_img.rows; r += step) {
+            for (int c = 0; c + 24 < resized_img.cols; c += step) {
+                eval->setWindow(Point(c, r));
+                if (predict(eval, &h) == +1) {
+                    int row = (int)(r / scale);
+                    int col = (int)(c / scale);
+                    bboxes.push_back(Rect(col, row, n_cols, n_rows));
+                    confidences.push_back(h);
+                }
+            }
+        }
+    }
+    std::vector<int> levels(bboxes.size(), 0);
+    groupRectangles(bboxes, levels, confidences, 3, 0.7);
+}
+
+void WaldBoost::fit(Mat& data_pos, Mat& data_neg)
+{
+    // data_pos: F x N_pos
+    // data_neg: F x N_neg
+    // every feature corresponds to row
+    // every sample corresponds to column
+    assert(data_pos.rows >= weak_count_);
+    assert(data_pos.rows == data_neg.rows);
+
+    std::vector<bool> feature_ignore;
+    for (int i = 0; i < data_pos.rows; ++i) {
+        feature_ignore.push_back(false);
+    }
+
+    Mat1f pos_weights(1, data_pos.cols, 1.0f / (2 * data_pos.cols));
+    Mat1f neg_weights(1, data_neg.cols, 1.0f / (2 * data_neg.cols));
+    Mat1f pos_trace(1, data_pos.cols, 0.0f);
+    Mat1f neg_trace(1, data_neg.cols, 0.0f);
+
+    bool quantize = false;
+    if (data_pos.type() != CV_8U) {
+        std::cerr << "quantize" << std::endl;
+        quantize = true;
+    }
+
+    Mat1f data_min, data_step;
+    int n_bins = 256;
+    if (quantize) {
+        compute_min_step(data_pos, data_neg, n_bins, data_min, data_step);
+        quantize_data(data_pos, data_min, data_step);
+        quantize_data(data_neg, data_min, data_step);
+    }
+
+    std::cerr << "pos=" << data_pos.cols << " neg=" << data_neg.cols << std::endl;
+    for (int i = 0; i < weak_count_; ++i) {
+        // Train weak learner with lowest error using weights
+        double min_err = DBL_MAX;
+        int min_feature_ind = -1;
+        int min_polarity = 0;
+        int threshold_q = 0;
+        float min_threshold = 0;
+//#pragma omp parallel for
+        for (int feat_i = 0; feat_i < data_pos.rows; ++feat_i) {
+            if (feature_ignore[feat_i])
+                continue;
+
+            // Construct cdf
+            Mat1f pos_cdf(1, n_bins), neg_cdf(1, n_bins);
+            compute_cdf(data_pos.row(feat_i), pos_weights, pos_cdf);
+            compute_cdf(data_neg.row(feat_i), neg_weights, neg_cdf);
+
+            float neg_total = (float)sum(neg_weights)[0];
+            Mat1f err_direct = pos_cdf + neg_total - neg_cdf;
+            Mat1f err_backward = 1.0f - err_direct;
+
+            int idx1[2], idx2[2];
+            double err1, err2;
+            minMaxIdx(err_direct, &err1, NULL, idx1);
+            minMaxIdx(err_backward, &err2, NULL, idx2);
+//#pragma omp critical
+            {
+            if (min(err1, err2) < min_err) {
+                if (err1 < err2) {
+                    min_err = err1;
+                    min_polarity = +1;
+                    threshold_q = idx1[1];
+                } else {
+                    min_err = err2;
+                    min_polarity = -1;
+                    threshold_q = idx2[1];
+                }
+                min_feature_ind = feat_i;
+                if (quantize) {
+                    min_threshold = data_min(feat_i, 0) + data_step(feat_i, 0) *
+                        (threshold_q + .5f);
+                } else {
+                    min_threshold = threshold_q + .5f;
+                }
+            }
+            }
+        }
+
+
+        float alpha = .5f * (float)log((1 - min_err) / min_err);
+        alphas_.push_back(alpha);
+        feature_indices_.push_back(min_feature_ind);
+        thresholds_.push_back(min_threshold);
+        polarities_.push_back(min_polarity);
+        feature_ignore[min_feature_ind] = true;
+
+        double loss = 0;
+        // Update positive weights
+        for (int j = 0; j < data_pos.cols; ++j) {
+            int val = data_pos.at<unsigned char>(min_feature_ind, j);
+            int label = min_polarity * (val - threshold_q) >= 0 ? +1 : -1;
+            pos_weights(0, j) *= exp(-alpha * label);
+            pos_trace(0, j) += alpha * label;
+            loss += exp(-pos_trace(0, j)) / (2.0f * data_pos.cols);
+        }
+
+        // Update negative weights
+        for (int j = 0; j < data_neg.cols; ++j) {
+            int val = data_neg.at<unsigned char>(min_feature_ind, j);
+            int label = min_polarity * (val - threshold_q) >= 0 ? +1 : -1;
+            neg_weights(0, j) *= exp(alpha * label);
+            neg_trace(0, j) += alpha * label;
+            loss += exp(+neg_trace(0, j)) / (2.0f * data_neg.cols);
+        }
+        double cascade_threshold = -1;
+        minMaxIdx(pos_trace, &cascade_threshold);
+        cascade_thresholds_.push_back((float)cascade_threshold);
+
+        std::cerr << "i=" << std::setw(4) << i;
+        std::cerr << " feat=" << std::setw(5) << min_feature_ind;
+        std::cerr << " thr=" << std::setw(3) << threshold_q;
+        std::cerr << " casthr=" << std::fixed << std::setprecision(3)
+             << cascade_threshold;
+        std::cerr <<  " alpha=" << std::fixed << std::setprecision(3)
+             << alpha << " err=" << std::fixed << std::setprecision(3) << min_err
+             << " loss=" << std::scientific << loss << std::endl;
+
+        //int pos = 0;
+        //for (int j = 0; j < data_pos.cols; ++j) {
+        //    if (pos_trace(0, j) > cascade_threshold - 0.5) {
+        //        pos_trace(0, pos) = pos_trace(0, j);
+        //        data_pos.col(j).copyTo(data_pos.col(pos));
+        //        pos_weights(0, pos) = pos_weights(0, j);
+        //        pos += 1;
+        //    }
+        //}
+        //std::cerr << "pos " << data_pos.cols << "/" << pos << std::endl;
+        //pos_trace = pos_trace.colRange(0, pos);
+        //data_pos = data_pos.colRange(0, pos);
+        //pos_weights = pos_weights.colRange(0, pos);
+
+        int pos = 0;
+        for (int j = 0; j < data_neg.cols; ++j) {
+            if (neg_trace(0, j) > cascade_threshold - 0.5) {
+                neg_trace(0, pos) = neg_trace(0, j);
+                data_neg.col(j).copyTo(data_neg.col(pos));
+                neg_weights(0, pos) = neg_weights(0, j);
+                pos += 1;
+            }
+        }
+        std::cerr << "neg " << data_neg.cols << "/" << pos << std::endl;
+        neg_trace = neg_trace.colRange(0, pos);
+        data_neg = data_neg.colRange(0, pos);
+        neg_weights = neg_weights.colRange(0, pos);
+
+
+        if (loss < 1e-50 || min_err > 0.5) {
+            std::cerr << "Stopping early. loss=" << loss << " min_err=" << min_err << std::endl;
+            weak_count_ = i + 1;
+            break;
+        }
+
+        // Avoid crashing on next Mat creation
+        if (pos <= 1) {
+            std::cerr << "Stopping early. pos=" << pos << std::endl;
+            weak_count_ = i + 1;
+            break;
+        }
+
+        // Normalize weights
+        double z = (sum(pos_weights) + sum(neg_weights))[0];
+        pos_weights /= z;
+        neg_weights /= z;
+    }
+}
+
+int WaldBoost::predict(Ptr<CvFeatureEvaluator> eval, float *h) const
+{
+    assert(feature_indices_.size() == size_t(weak_count_));
+    assert(cascade_thresholds_.size() == size_t(weak_count_));
+    float res = 0;
+    int count = weak_count_;
+    for (int i = 0; i < count; ++i) {
+        float val = (*eval)(feature_indices_[i]);
+        int label = polarities_[i] * (val - thresholds_[i]) > 0 ? +1: -1;
+        res += alphas_[i] * label;
+        if (res < cascade_thresholds_[i]) {
+            return -1;
+        }
+    }
+    *h = res;
+    return res > cascade_thresholds_[count - 1] ? +1 : -1;
+}
+
+void WaldBoost::write(FileStorage &fs) const
+{
+    fs << "{";
+    fs << "waldboost_params"
+       << "{" << "weak_count" << weak_count_ << "}";
+
+    fs << "thresholds" << "[";
+    for (size_t i = 0; i < thresholds_.size(); ++i)
+        fs << thresholds_[i];
+    fs << "]";
+
+    fs << "alphas" << "[";
+    for (size_t i = 0; i < alphas_.size(); ++i)
+        fs << alphas_[i];
+    fs << "]";
+
+    fs << "polarities" << "[";
+    for (size_t i = 0; i < polarities_.size(); ++i)
+        fs << polarities_[i];
+    fs << "]";
+
+    fs << "cascade_thresholds" << "[";
+    for (size_t i = 0; i < cascade_thresholds_.size(); ++i)
+        fs << cascade_thresholds_[i];
+    fs << "]";
+
+    fs << "feature_indices" << "[";
+    for (size_t i = 0; i < feature_indices_.size(); ++i)
+        fs << feature_indices_[i];
+    fs << "]";
+
+    fs << "}";
+}
+
+void WaldBoost::read(const FileNode &node)
+{
+    weak_count_ = (int)(node["waldboost_params"]["weak_count"]);
+    thresholds_.resize(weak_count_);
+    alphas_.resize(weak_count_);
+    polarities_.resize(weak_count_);
+    cascade_thresholds_.resize(weak_count_);
+    feature_indices_.resize(weak_count_);
+
+    FileNodeIterator n;
+
+    n = node["thresholds"].begin();
+    for (int i = 0; i < weak_count_; ++i, ++n)
+        *n >> thresholds_[i];
+
+    n = node["alphas"].begin();
+    for (int i = 0; i < weak_count_; ++i, ++n)
+        *n >> alphas_[i];
+
+    n = node["polarities"].begin();
+    for (int i = 0; i < weak_count_; ++i, ++n)
+        *n >> polarities_[i];
+
+    n = node["cascade_thresholds"].begin();
+    for (int i = 0; i < weak_count_; ++i, ++n)
+        *n >> cascade_thresholds_[i];
+
+    n = node["feature_indices"].begin();
+    for (int i = 0; i < weak_count_; ++i, ++n)
+        *n >> feature_indices_[i];
+}
+
+void WaldBoost::reset(int weak_count)
+{
+    weak_count_ = weak_count;
+    thresholds_.clear();
+    alphas_.clear();
+    feature_indices_.clear();
+    polarities_.clear();
+    cascade_thresholds_.clear();
+}
+
+WaldBoost::~WaldBoost()
+{
+}
+
+}
+}

modules/xobjdetect/src/waldboost.hpp

@@ -0,0 +1,94 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#ifndef __OPENCV_XOBJDETECT_WALDBOOST_HPP__
+#define __OPENCV_XOBJDETECT_WALDBOOST_HPP__
+
+#include "precomp.hpp"
+
+namespace cv {
+namespace xobjdetect {
+
+class WaldBoost {
+public:
+    WaldBoost(int weak_count);
+    WaldBoost();
+    std::vector<int> get_feature_indices();
+
+    void detect(Ptr<CvFeatureEvaluator> eval,
+                const Mat& img,
+                const std::vector<float>& scales,
+                std::vector<Rect>& bboxes,
+                Mat1f& confidences);
+
+    void detect(Ptr<CvFeatureEvaluator> eval,
+                const Mat& img,
+                const std::vector<float>& scales,
+                std::vector<Rect>& bboxes,
+                std::vector<double>& confidences);
+
+    void fit(Mat& data_pos, Mat& data_neg);
+    int predict(Ptr<CvFeatureEvaluator> eval, float *h) const;
+    void save(const std::string& filename);
+    void load(const std::string& filename);
+
+    void read(const FileNode &node);
+    void write(FileStorage &fs) const;
+
+    void reset(int weak_count);
+    ~WaldBoost();
+
+private:
+    int weak_count_;
+    std::vector<float> thresholds_;
+    std::vector<float> alphas_;
+    std::vector<int> feature_indices_;
+    std::vector<int> polarities_;
+    std::vector<float> cascade_thresholds_;
+};
+
+}
+}
+
+#endif

modules/xobjdetect/src/wbdetector.cpp

@@ -0,0 +1,222 @@
+/*
+By downloading, copying, installing or using the software you agree to this license.
+If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2000-2015, Intel Corporation, all rights reserved.
+Copyright (C) 2009-2011, Willow Garage Inc., all rights reserved.
+Copyright (C) 2009-2015, NVIDIA Corporation, all rights reserved.
+Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+Copyright (C) 2015, OpenCV Foundation, all rights reserved.
+Copyright (C) 2015, Itseez Inc., all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are disclaimed.
+In no event shall copyright holders or contributors be liable for any direct,
+indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+*/
+
+#include "precomp.hpp"
+
+using std::cerr;
+using std::endl;
+using std::vector;
+using std::string;
+
+namespace cv {
+namespace xobjdetect {
+
+static vector<Mat> sample_patches(
+        const string& path,
+        int n_rows,
+        int n_cols,
+        size_t n_patches)
+{
+    vector<String> filenames;
+    glob(path, filenames);
+    vector<Mat> patches;
+    size_t patch_count = 0;
+    for (size_t i = 0; i < filenames.size(); ++i) {
+        Mat img = imread(filenames[i], IMREAD_GRAYSCALE);
+        for (int row = 0; row + n_rows < img.rows; row += n_rows) {
+            for (int col = 0; col + n_cols < img.cols; col += n_cols) {
+                patches.push_back(img(Rect(col, row, n_cols, n_rows)).clone());
+                patch_count += 1;
+                if (patch_count == n_patches) {
+                    goto sampling_finished;
+                }
+            }
+        }
+    }
+sampling_finished:
+    return patches;
+}
+
+static vector<Mat> read_imgs(const string& path)
+{
+    vector<String> filenames;
+    glob(path, filenames);
+    vector<Mat> imgs;
+    for (size_t i = 0; i < filenames.size(); ++i) {
+        imgs.push_back(imread(filenames[i], IMREAD_GRAYSCALE));
+    }
+    return imgs;
+}
+
+void WBDetectorImpl::read(const FileNode& node)
+{
+    boost_.read(node);
+}
+
+
+void WBDetectorImpl::write(FileStorage &fs) const
+{
+    boost_.write(fs);
+}
+
+void WBDetectorImpl::train(
+    const string& pos_samples_path,
+    const string& neg_imgs_path)
+{
+
+    vector<Mat> pos_imgs = read_imgs(pos_samples_path);
+    vector<Mat> neg_imgs = sample_patches(neg_imgs_path, 24, 24, pos_imgs.size() * 10);
+
+    CV_Assert(pos_imgs.size());
+    CV_Assert(neg_imgs.size());
+
+    int n_features;
+    Mat pos_data, neg_data;
+
+    Ptr<CvFeatureEvaluator> eval = CvFeatureEvaluator::create();
+    eval->init(CvFeatureParams::create(), 1, Size(24, 24));
+    n_features = eval->getNumFeatures();
+
+    const int stages[] = {64, 128, 256, 512, 1024};
+    const int stage_count = sizeof(stages) / sizeof(*stages);
+    const int stage_neg = (int)(pos_imgs.size() * 5);
+    const int max_per_image = 100;
+
+    const float scales_arr[] = {.3f, .4f, .5f, .6f, .7f, .8f, .9f, 1.0f};
+    const vector<float> scales(scales_arr,
+            scales_arr + sizeof(scales_arr) / sizeof(*scales_arr));
+
+    vector<String> neg_filenames;
+    glob(neg_imgs_path, neg_filenames);
+
+
+    for (int i = 0; i < stage_count; ++i) {
+
+        cerr << "compute features" << endl;
+
+        pos_data = Mat1b(n_features, (int)pos_imgs.size());
+        neg_data = Mat1b(n_features, (int)neg_imgs.size());
+
+        for (size_t k = 0; k < pos_imgs.size(); ++k) {
+            eval->setImage(pos_imgs[k], +1, 0, boost_.get_feature_indices());
+            for (int j = 0; j < n_features; ++j) {
+                pos_data.at<uchar>(j, (int)k) = (uchar)(*eval)(j);
+            }
+        }
+
+        for (size_t k = 0; k < neg_imgs.size(); ++k) {
+            eval->setImage(neg_imgs[k], 0, 0, boost_.get_feature_indices());
+            for (int j = 0; j < n_features; ++j) {
+                neg_data.at<uchar>(j, (int)k) = (uchar)(*eval)(j);
+            }
+        }
+
+
+        boost_.reset(stages[i]);
+        boost_.fit(pos_data, neg_data);
+
+        if (i + 1 == stage_count) {
+            break;
+        }
+
+        int bootstrap_count = 0;
+        size_t img_i = 0;
+        for (; img_i < neg_filenames.size(); ++img_i) {
+            cerr << "win " << bootstrap_count << "/" << stage_neg
+                 << " img " << (img_i + 1) << "/" << neg_filenames.size() << "\r";
+            Mat img = imread(neg_filenames[img_i], IMREAD_GRAYSCALE);
+            vector<Rect> bboxes;
+            Mat1f confidences;
+            boost_.detect(eval, img, scales, bboxes, confidences);
+
+            if (confidences.rows > 0) {
+                Mat1i indices;
+                sortIdx(confidences, indices,
+                        SORT_EVERY_COLUMN + SORT_DESCENDING);
+
+                int win_count = min(max_per_image, confidences.rows);
+                win_count = min(win_count, stage_neg - bootstrap_count);
+                Mat window;
+                for (int k = 0; k < win_count; ++k) {
+                    resize(img(bboxes[indices(k, 0)]), window, Size(24, 24), 0, 0, INTER_LINEAR_EXACT);
+                    neg_imgs.push_back(window.clone());
+                    bootstrap_count += 1;
+                }
+                if (bootstrap_count >= stage_neg) {
+                    break;
+                }
+            }
+        }
+        cerr << "bootstrapped " << bootstrap_count << " windows from "
+             << (img_i + 1) << " images" << endl;
+    }
+}
+
+void WBDetectorImpl::detect(
+    const Mat& img,
+    vector<Rect> &bboxes,
+    vector<double> &confidences)
+{
+    Mat test_img = img.clone();
+    bboxes.clear();
+    confidences.clear();
+    vector<float> scales;
+    for (float scale = 0.2f; scale < 1.2f; scale *= 1.1f) {
+        scales.push_back(scale);
+    }
+    Ptr<CvFeatureParams> params = CvFeatureParams::create();
+    Ptr<CvFeatureEvaluator> eval = CvFeatureEvaluator::create();
+    eval->init(params, 1, Size(24, 24));
+    boost_.detect(eval, img, scales, bboxes, confidences);
+    CV_Assert(confidences.size() == bboxes.size());
+}
+
+Ptr<WBDetector>
+WBDetector::create()
+{
+    return Ptr<WBDetector>(new WBDetectorImpl());
+}
+
+}
+}

modules/xobjdetect/src/wbdetector.hpp

@@ -0,0 +1,76 @@
+/*
+
+By downloading, copying, installing or using the software you agree to this
+license. If you do not agree to this license, do not download, install,
+copy or use the software.
+
+
+                          License Agreement
+               For Open Source Computer Vision Library
+                       (3-clause BSD License)
+
+Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+Third party copyrights are property of their respective owners.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+  * Redistributions of source code must retain the above copyright notice,
+    this list of conditions and the following disclaimer.
+
+  * Redistributions in binary form must reproduce the above copyright notice,
+    this list of conditions and the following disclaimer in the documentation
+    and/or other materials provided with the distribution.
+
+  * Neither the names of the copyright holders nor the names of the contributors
+    may be used to endorse or promote products derived from this software
+    without specific prior written permission.
+
+This software is provided by the copyright holders and contributors "as is" and
+any express or implied warranties, including, but not limited to, the implied
+warranties of merchantability and fitness for a particular purpose are
+disclaimed. In no event shall copyright holders or contributors be liable for
+any direct, indirect, incidental, special, exemplary, or consequential damages
+(including, but not limited to, procurement of substitute goods or services;
+loss of use, data, or profits; or business interruption) however caused
+and on any theory of liability, whether in contract, strict liability,
+or tort (including negligence or otherwise) arising in any way out of
+the use of this software, even if advised of the possibility of such damage.
+
+*/
+
+#ifndef __OPENCV_XOBJDETECT_DETECTOR_HPP__
+#define __OPENCV_XOBJDETECT_DETECTOR_HPP__
+
+#include <opencv2/core.hpp>
+#include <vector>
+#include <string>
+#include "precomp.hpp"
+
+namespace cv
+{
+namespace xobjdetect
+{
+
+class WBDetectorImpl CV_FINAL : public WBDetector {
+public:
+    virtual void read(const FileNode &node) CV_OVERRIDE;
+    virtual void write(FileStorage &fs) const CV_OVERRIDE;
+
+    virtual void train(
+        const std::string& pos_samples,
+        const std::string& neg_imgs) CV_OVERRIDE;
+
+    virtual void detect(
+        const Mat& img,
+        std::vector<Rect> &bboxes,
+        std::vector<double> &confidences) CV_OVERRIDE;
+
+private:
+    WaldBoost boost_;
+};
+
+} /* namespace xobjdetect */
+} /* namespace cv */
+
+#endif /* __OPENCV_XOBJDETECT_DETECTOR_HPP__ */

modules/xobjdetect/test/opencl/test_hogdetector.cpp

@@ -0,0 +1,140 @@
+///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
+// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
+// Copyright (C) 2010-2012, Multicoreware, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// @Authors
+//    Niko Li, newlife20080214@gmail.com
+//    Jia Haipeng, jiahaipeng95@gmail.com
+//    Shengen Yan, yanshengen@gmail.com
+//    Jiang Liyuan,jlyuan001.good@163.com
+//    Rock Li, Rock.Li@amd.com
+//    Zailong Wu, bullet@yeah.net
+//    Yao Wang, bitwangyaoyao@gmail.com
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "../test_precomp.hpp"
+#include "opencv2/ts/ocl_test.hpp"
+
+#ifdef HAVE_OPENCL
+
+namespace opencv_test {
+namespace ocl {
+
+///////////////////// HOG /////////////////////////////
+PARAM_TEST_CASE(HOG, Size, MatType)
+{
+    Size winSize;
+    int type;
+    Mat img;
+    UMat uimg;
+    virtual void SetUp()
+    {
+        winSize = GET_PARAM(0);
+        type = GET_PARAM(1);
+        img = readImage("cascadeandhog/images/image_00000000_0.png", IMREAD_GRAYSCALE);
+        ASSERT_FALSE(img.empty());
+        img.copyTo(uimg);
+    }
+};
+
+OCL_TEST_P(HOG, GetDescriptors)
+{
+    HOGDescriptor hog;
+    hog.gammaCorrection = true;
+
+    hog.setSVMDetector(hog.getDefaultPeopleDetector());
+
+    std::vector<float> cpu_descriptors;
+    std::vector<float> gpu_descriptors;
+
+    OCL_OFF(hog.compute(img, cpu_descriptors, hog.winSize));
+    OCL_ON(hog.compute(uimg, gpu_descriptors, hog.winSize));
+
+    Mat cpu_desc(cpu_descriptors), gpu_desc(gpu_descriptors);
+
+    EXPECT_MAT_SIMILAR(cpu_desc, gpu_desc, 1e-1);
+}
+
+OCL_TEST_P(HOG, SVMDetector)
+{
+    HOGDescriptor hog_first, hog_second;
+
+    // empty -> empty
+    hog_first.copyTo(hog_second);
+
+    // first -> both
+    hog_first.setSVMDetector(hog_first.getDefaultPeopleDetector());
+    hog_first.copyTo(hog_second);
+
+    // both -> both
+    hog_first.copyTo(hog_second);
+
+    // second -> empty
+    hog_first.setSVMDetector(cv::noArray());
+    hog_first.copyTo(hog_second);
+}
+
+OCL_TEST_P(HOG, Detect)
+{
+    HOGDescriptor hog;
+    hog.winSize = winSize;
+    hog.gammaCorrection = true;
+
+    if (winSize.width == 48 && winSize.height == 96)
+        hog.setSVMDetector(hog.getDaimlerPeopleDetector());
+    else
+        hog.setSVMDetector(hog.getDefaultPeopleDetector());
+
+    std::vector<Rect> cpu_found;
+    std::vector<Rect> gpu_found;
+
+    OCL_OFF(hog.detectMultiScale(img, cpu_found, 0, Size(8, 8), Size(0, 0), 1.05, 6));
+    OCL_ON(hog.detectMultiScale(uimg, gpu_found, 0, Size(8, 8), Size(0, 0), 1.05, 6));
+
+    EXPECT_LT(checkRectSimilarity(img.size(), cpu_found, gpu_found), 0.05);
+}
+
+INSTANTIATE_TEST_CASE_P(OCL_ObjDetect, HOG, testing::Combine(
+                            testing::Values(Size(64, 128), Size(48, 96)),
+                            testing::Values( MatType(CV_8UC1) ) ) );
+
+}} // namespace
+#endif

modules/xobjdetect/test/test_main.cpp

@@ -0,0 +1,10 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+#include "test_precomp.hpp"
+
+#if defined(HAVE_HPX)
+    #include <hpx/hpx_main.hpp>
+#endif
+
+CV_TEST_MAIN("cv")

modules/xobjdetect/test/test_precomp.hpp

@@ -0,0 +1,12 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+#ifndef __OPENCV_TEST_PRECOMP_HPP__
+#define __OPENCV_TEST_PRECOMP_HPP__
+
+#include "opencv2/ts.hpp"
+#include "opencv2/xobjdetect.hpp"
+
+#include <random>
+
+#endif

modules/xobjdetect/tools/CMakeLists.txt

@@ -0,0 +1,35 @@
+function(ocv_add_tool the_target)
+  cmake_parse_arguments(${the_target} "" "" "MODULES" ${ARGN})
+  ocv_check_dependencies(${${the_target}_MODULES})
+  if(NOT OCV_DEPENDENCIES_FOUND)
+    return()
+  endif()
+
+  project(${the_target})
+
+  ocv_include_directories(${the_target} PRIVATE "${OpenCV_SOURCE_DIR}/include/opencv")
+  ocv_include_modules_recurse(${the_target} ${${the_target}_MODULES})
+
+  file(GLOB ${the_target}_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/*.cpp)
+
+  add_executable(${the_target} ${${the_target}_SOURCES})
+
+  ocv_target_link_libraries(${the_target} ${${the_target}_MODULES})
+
+  set_target_properties(${the_target} PROPERTIES
+                          DEBUG_POSTFIX "${OPENCV_DEBUG_POSTFIX}"
+                          ARCHIVE_OUTPUT_DIRECTORY ${LIBRARY_OUTPUT_PATH}
+                          RUNTIME_OUTPUT_DIRECTORY ${EXECUTABLE_OUTPUT_PATH}
+                          OUTPUT_NAME ${the_target})
+
+  if(ENABLE_SOLUTION_FOLDERS)
+    set_target_properties(${the_target} PROPERTIES FOLDER "applications")
+  endif()
+
+  install(TARGETS ${the_target} OPTIONAL RUNTIME DESTINATION bin COMPONENT main)
+endfunction()
+
+
+add_subdirectory(waldboost_detector)
+add_subdirectory(annotation)
+add_subdirectory(visualisation)

modules/xobjdetect/tools/annotation/CMakeLists.txt

@@ -0,0 +1 @@
+ocv_add_tool(opencv_annotation MODULES opencv_core opencv_highgui opencv_imgproc opencv_imgcodecs opencv_videoio)

modules/xobjdetect/tools/annotation/opencv_annotation.cpp

@@ -0,0 +1,317 @@
+////////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+////////////////////////////////////////////////////////////////////////////////////////
+
+/*****************************************************************************************************
+USAGE:
+./opencv_annotation -images <folder location> -annotations <output file>
+
+Created by: Puttemans Steven - February 2015
+Adapted by: Puttemans Steven - April 2016 - Vectorize the process to enable better processing
+                                               + early leave and store by pressing an ESC key
+                                               + enable delete `d` button, to remove last annotation
+*****************************************************************************************************/
+
+#include <opencv2/core.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/imgcodecs.hpp>
+#include <opencv2/videoio.hpp>
+#include <opencv2/imgproc.hpp>
+
+#include <fstream>
+#include <iostream>
+#include <map>
+
+using namespace std;
+using namespace cv;
+
+// Function prototypes
+void on_mouse(int, int, int, int, void*);
+vector<Rect> get_annotations(Mat);
+
+// Public parameters
+Mat image;
+int roi_x0 = 0, roi_y0 = 0, roi_x1 = 0, roi_y1 = 0, num_of_rec = 0;
+bool start_draw = false, stop = false;
+
+// Window name for visualisation purposes
+const string window_name = "OpenCV Based Annotation Tool";
+
+// FUNCTION : Mouse response for selecting objects in images
+// If left button is clicked, start drawing a rectangle as long as mouse moves
+// Stop drawing once a new left click is detected by the on_mouse function
+void on_mouse(int event, int x, int y, int , void * )
+{
+    // Action when left button is clicked
+    if(event == EVENT_LBUTTONDOWN)
+    {
+        if(!start_draw)
+        {
+            roi_x0 = x;
+            roi_y0 = y;
+            start_draw = true;
+        } else {
+            roi_x1 = x;
+            roi_y1 = y;
+            start_draw = false;
+        }
+    }
+
+    // Action when mouse is moving and drawing is enabled
+    if((event == EVENT_MOUSEMOVE) && start_draw)
+    {
+        // Redraw bounding box for annotation
+        Mat current_view;
+        image.copyTo(current_view);
+        rectangle(current_view, Point(roi_x0,roi_y0), Point(x,y), Scalar(0,0,255));
+        imshow(window_name, current_view);
+    }
+}
+
+// FUNCTION : returns a vector of Rect objects given an image containing positive object instances
+vector<Rect> get_annotations(Mat input_image)
+{
+    vector<Rect> current_annotations;
+
+    // Make it possible to exit the annotation process
+    stop = false;
+
+    // Init window interface and couple mouse actions
+    namedWindow(window_name, WINDOW_AUTOSIZE);
+    setMouseCallback(window_name, on_mouse);
+
+    image = input_image;
+    imshow(window_name, image);
+    int key_pressed = 0;
+
+    do
+    {
+        // Get a temporary image clone
+        Mat temp_image = input_image.clone();
+        Rect currentRect(0, 0, 0, 0);
+
+        // Keys for processing
+        // You need to select one for confirming a selection and one to continue to the next image
+        // Based on the universal ASCII code of the keystroke: http://www.asciitable.com/
+        //      c = 99		    add rectangle to current image
+        //	    n = 110		    save added rectangles and show next image
+        //      d = 100         delete the last annotation made
+        //	    <ESC> = 27      exit program
+        key_pressed = 0xFF & waitKey(0);
+        switch( key_pressed )
+        {
+        case 27:
+                stop = true;
+                break;
+        case 99:
+                // Draw initiated from top left corner
+                if(roi_x0<roi_x1 && roi_y0<roi_y1)
+                {
+                    currentRect.x = roi_x0;
+                    currentRect.y = roi_y0;
+                    currentRect.width = roi_x1-roi_x0;
+                    currentRect.height = roi_y1-roi_y0;
+                }
+                // Draw initiated from bottom right corner
+                if(roi_x0>roi_x1 && roi_y0>roi_y1)
+                {
+                    currentRect.x = roi_x1;
+                    currentRect.y = roi_y1;
+                    currentRect.width = roi_x0-roi_x1;
+                    currentRect.height = roi_y0-roi_y1;
+                }
+                // Draw initiated from top right corner
+                if(roi_x0>roi_x1 && roi_y0<roi_y1)
+                {
+                    currentRect.x = roi_x1;
+                    currentRect.y = roi_y0;
+                    currentRect.width = roi_x0-roi_x1;
+                    currentRect.height = roi_y1-roi_y0;
+                }
+                // Draw initiated from bottom left corner
+                if(roi_x0<roi_x1 && roi_y0>roi_y1)
+                {
+                    currentRect.x = roi_x0;
+                    currentRect.y = roi_y1;
+                    currentRect.width = roi_x1-roi_x0;
+                    currentRect.height = roi_y0-roi_y1;
+                }
+                // Draw the rectangle on the canvas
+                // Add the rectangle to the vector of annotations
+                current_annotations.push_back(currentRect);
+                break;
+        case 100:
+                // Remove the last annotation
+                if(current_annotations.size() > 0){
+                    current_annotations.pop_back();
+                }
+                break;
+        default:
+                // Default case --> do nothing at all
+                // Other keystrokes can simply be ignored
+                break;
+        }
+
+        // Check if escape has been pressed
+        if(stop)
+        {
+            break;
+        }
+
+        // Draw all the current rectangles onto the top image and make sure that the global image is linked
+        for(int i=0; i < (int)current_annotations.size(); i++){
+            rectangle(temp_image, current_annotations[i], Scalar(0,255,0), 1);
+        }
+        image = temp_image;
+
+        // Force an explicit redraw of the canvas --> necessary to visualize delete correctly
+        imshow(window_name, image);
+    }
+    // Continue as long as the next image key has not been pressed
+    while(key_pressed != 110);
+
+    // Close down the window
+    destroyWindow(window_name);
+
+    // Return the data
+    return current_annotations;
+}
+
+int main( int argc, const char** argv )
+{
+    // Use the cmdlineparser to process input arguments
+    CommandLineParser parser(argc, argv,
+        "{ help h usage ? |      | show this message }"
+        "{ images i       |      | (required) path to image folder [example - /data/testimages/] }"
+        "{ annotations a  |      | (required) path to annotations txt file [example - /data/annotations.txt] }"
+        "{ maxWindowHeight m  |  -1   | (optional) images larger in height than this value will be scaled down }"
+        "{ resizeFactor r  |  2  | (optional) factor for scaling down [default = half the size] }"
+    );
+    // Read in the input arguments
+    if (parser.has("help")){
+        parser.printMessage();
+        cerr << "TIP: Use absolute paths to avoid any problems with the software!" << endl;
+        return 0;
+    }
+    string image_folder(parser.get<string>("images"));
+    string annotations_file(parser.get<string>("annotations"));
+    if (image_folder.empty() || annotations_file.empty()){
+        parser.printMessage();
+        cerr << "TIP: Use absolute paths to avoid any problems with the software!" << endl;
+        return -1;
+    }
+
+    int resizeFactor = parser.get<int>("resizeFactor");
+    int const maxWindowHeight = parser.get<int>("maxWindowHeight") > 0 ? parser.get<int>("maxWindowHeight") : -1;
+
+    // Start by processing the data
+    // Return the image filenames inside the image folder
+    map< String, vector<Rect> > annotations;
+    vector<String> filenames;
+    String folder(image_folder);
+    glob(folder, filenames);
+
+    // Add key tips on how to use the software when running it
+    cout << "* mark rectangles with the left mouse button," << endl;
+    cout << "* press 'c' to accept a selection," << endl;
+    cout << "* press 'd' to delete the latest selection," << endl;
+    cout << "* press 'n' to proceed with next image," << endl;
+    cout << "* press 'esc' to stop." << endl;
+
+    // Loop through each image stored in the images folder
+    // Create and temporarily store the annotations
+    // At the end write everything to the annotations file
+    for (size_t i = 0; i < filenames.size(); i++){
+        // Read in an image
+        Mat current_image = imread(filenames[i]);
+        bool const resize_bool = (maxWindowHeight > 0) && (current_image.rows > maxWindowHeight);
+
+        // Check if the image is actually read - avoid other files in the folder, because glob() takes them all
+        // If not then simply skip this iteration
+        if(current_image.empty()){
+            continue;
+        }
+
+        if(resize_bool){
+            resize(current_image, current_image, Size(current_image.cols/resizeFactor, current_image.rows/resizeFactor), 0, 0, INTER_LINEAR_EXACT);
+        }
+
+        // Perform annotations & store the result inside the vectorized structure
+        // If the image was resized before, then resize the found annotations back to original dimensions
+        vector<Rect> current_annotations = get_annotations(current_image);
+        if(resize_bool){
+            for(int j =0; j < (int)current_annotations.size(); j++){
+                current_annotations[j].x = current_annotations[j].x * resizeFactor;
+                current_annotations[j].y = current_annotations[j].y * resizeFactor;
+                current_annotations[j].width = current_annotations[j].width * resizeFactor;
+                current_annotations[j].height = current_annotations[j].height * resizeFactor;
+            }
+        }
+        annotations[filenames[i]] = current_annotations;
+
+        // Check if the ESC key was hit, then exit earlier then expected
+        if(stop){
+            break;
+        }
+    }
+
+    // When all data is processed, store the data gathered inside the proper file
+    // This now even gets called when the ESC button was hit to store preliminary results
+    ofstream output(annotations_file.c_str());
+    if ( !output.is_open() ){
+        cerr << "The path for the output file contains an error and could not be opened. Please check again!" << endl;
+        return 0;
+    }
+
+    // Store the annotations, write to the output file
+    for(map<String, vector<Rect> >::iterator it = annotations.begin(); it != annotations.end(); it++){
+        vector<Rect> &anno = it->second;
+        output << it->first << " " << anno.size();
+        for(size_t j=0; j < anno.size(); j++){
+            Rect temp = anno[j];
+            output << " " << temp.x << " " << temp.y << " " << temp.width << " " << temp.height;
+        }
+        output << endl;
+    }
+
+    return 0;
+}

modules/xobjdetect/tools/visualisation/CMakeLists.txt

@@ -0,0 +1 @@
+ocv_add_tool(opencv_visualisation MODULES opencv_core opencv_highgui opencv_imgproc opencv_videoio opencv_imgcodecs)

modules/xobjdetect/tools/visualisation/opencv_visualisation.cpp

@@ -0,0 +1,378 @@
+////////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+////////////////////////////////////////////////////////////////////////////////////////
+
+/*****************************************************************************************************
+
+Software for visualising cascade classifier models trained by OpenCV and to get a better
+understanding of the used features.
+
+USAGE:
+./opencv_visualisation --model=<model.xml> --image=<ref.png> --data=<video output folder>
+
+Created by: Puttemans Steven - April 2016
+*****************************************************************************************************/
+
+#include <opencv2/core.hpp>
+#include <opencv2/highgui.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/imgcodecs.hpp>
+#include <opencv2/videoio.hpp>
+
+#include <fstream>
+#include <iostream>
+#include <sstream>
+
+using namespace std;
+using namespace cv;
+
+struct rect_data{
+    int x;
+    int y;
+    int w;
+    int h;
+    float weight;
+};
+
+static void printLimits(){
+    cerr << "Limits of the current interface:" << endl;
+    cerr << " - Only handles cascade classifier models, trained with the opencv_traincascade tool, containing stumps as decision trees [default settings]." << endl;
+    cerr << " - The image provided needs to be a sample window with the original model dimensions, passed to the --image parameter." << endl;
+    cerr << " - ONLY handles HAAR and LBP features." << endl;
+}
+
+int main( int argc, const char** argv )
+{
+    CommandLineParser parser(argc, argv,
+        "{ help h usage ? |      | show this message }"
+        "{ image i        |      | (required) path to reference image }"
+        "{ model m        |      | (required) path to cascade xml file }"
+        "{ data d         |      | (optional) path to video output folder }"
+        "{ ext            | avi  | (optional) output video file extension e.g. avi (default) or mp4 }"
+        "{ fourcc         | XVID | (optional) output video file's 4-character codec e.g. XVID (default) or H264 }"
+        "{ fps            |   15 | (optional) output video file's frames-per-second rate }"
+    );
+    // Read in the input arguments
+    if (parser.has("help")){
+        parser.printMessage();
+        printLimits();
+        return 0;
+    }
+    string model(parser.get<string>("model"));
+    string output_folder(parser.get<string>("data"));
+    string image_ref = (parser.get<string>("image"));
+    string fourcc = (parser.get<string>("fourcc"));
+    int fps = parser.get<int>("fps");
+    if (model.empty() || image_ref.empty() || fourcc.size()!=4 || fps<1){
+        parser.printMessage();
+        printLimits();
+        return -1;
+    }
+
+    // Value for timing
+    // You can increase this to have a better visualisation during the generation
+    int timing = 1;
+
+    // Value for cols of storing elements
+    int cols_prefered = 5;
+
+    // Open the XML model
+    FileStorage fs;
+    bool model_ok = fs.open(model, FileStorage::READ);
+    if (!model_ok){
+        cerr << "the cascade file '" << model << "' could not be loaded." << endl;
+        return  -1;
+    }
+    // Get a the required information
+    // First decide which feature type we are using
+    FileNode cascade = fs["cascade"];
+    string feature_type = cascade["featureType"];
+    bool haar = false, lbp = false;
+    if (feature_type.compare("HAAR") == 0){
+        haar = true;
+    }
+    if (feature_type.compare("LBP") == 0){
+        lbp = true;
+    }
+    if ( feature_type.compare("HAAR") != 0 && feature_type.compare("LBP")){
+        cerr << "The model is not an HAAR or LBP feature based model!" << endl;
+        cerr << "Please select a model that can be visualized by the software." << endl;
+        return -1;
+    }
+
+    // We make a visualisation mask - which increases the window to make it at least a bit more visible
+    int resize_factor = 10;
+    int resize_storage_factor = 10;
+    Mat reference_image = imread(image_ref, IMREAD_GRAYSCALE );
+    if (reference_image.empty()){
+        cerr << "the reference image '" << image_ref << "'' could not be loaded." << endl;
+        return -1;
+    }
+    Mat visualization;
+    resize(reference_image, visualization, Size(reference_image.cols * resize_factor, reference_image.rows * resize_factor), 0, 0, INTER_LINEAR_EXACT);
+
+    // First recover for each stage the number of weak features and their index
+    // Important since it is NOT sequential when using LBP features
+    vector< vector<int> > stage_features;
+    FileNode stages = cascade["stages"];
+    FileNodeIterator it_stages = stages.begin(), it_stages_end = stages.end();
+    int idx = 0;
+    for( ; it_stages != it_stages_end; it_stages++, idx++ ){
+        vector<int> current_feature_indexes;
+        FileNode weak_classifiers = (*it_stages)["weakClassifiers"];
+        FileNodeIterator it_weak = weak_classifiers.begin(), it_weak_end = weak_classifiers.end();
+        vector<int> values;
+        for(int idy = 0; it_weak != it_weak_end; it_weak++, idy++ ){
+            (*it_weak)["internalNodes"] >> values;
+            current_feature_indexes.push_back( (int)values[2] );
+        }
+        stage_features.push_back(current_feature_indexes);
+    }
+
+    // If the output option has been chosen than we will store a combined image plane for
+    // each stage, containing all weak classifiers for that stage.
+    bool draw_planes = false;
+    stringstream output_video;
+    output_video << output_folder << "model_visualization." << parser.get<string>("ext");
+    VideoWriter result_video;
+    if( output_folder.compare("") != 0 ){
+        draw_planes = true;
+        result_video.open(output_video.str(), VideoWriter::fourcc(fourcc[0],fourcc[1],fourcc[2],fourcc[3]), fps, visualization.size(), false);
+        if (!result_video.isOpened()){
+            cerr << "the output video '" << output_video.str() << "' could not be opened."
+                 << " fourcc=" << fourcc
+                 << " fps=" << fps
+                 << " frameSize=" << visualization.size()
+                 << endl;
+            return -1;
+        }
+    }
+
+    if(haar){
+        // Grab the corresponding features dimensions and weights
+        FileNode features = cascade["features"];
+        vector< vector< rect_data > > feature_data;
+        FileNodeIterator it_features = features.begin(), it_features_end = features.end();
+        for(int idf = 0; it_features != it_features_end; it_features++, idf++ ){
+            vector< rect_data > current_feature_rectangles;
+            FileNode rectangles = (*it_features)["rects"];
+            int nrects = (int)rectangles.size();
+            for(int k = 0; k < nrects; k++){
+                rect_data current_data;
+                FileNode single_rect = rectangles[k];
+                current_data.x = (int)single_rect[0];
+                current_data.y = (int)single_rect[1];
+                current_data.w = (int)single_rect[2];
+                current_data.h = (int)single_rect[3];
+                current_data.weight = (float)single_rect[4];
+                current_feature_rectangles.push_back(current_data);
+            }
+            feature_data.push_back(current_feature_rectangles);
+        }
+
+        // Loop over each possible feature on its index, visualise on the mask and wait a bit,
+        // then continue to the next feature.
+        // If visualisations should be stored then do the in between calculations
+        Mat image_plane;
+        Mat metadata = Mat::zeros(150, 1000, CV_8UC1);
+        vector< rect_data > current_rects;
+        for(int sid = 0; sid < (int)stage_features.size(); sid ++){
+            if(draw_planes){
+                int features_nmbr = (int)stage_features[sid].size();
+                int cols = cols_prefered;
+                int rows = features_nmbr / cols;
+                if( (features_nmbr % cols) > 0){
+                    rows++;
+                }
+                image_plane = Mat::zeros(reference_image.rows * resize_storage_factor * rows, reference_image.cols * resize_storage_factor * cols, CV_8UC1);
+            }
+            for(int fid = 0; fid < (int)stage_features[sid].size(); fid++){
+                stringstream meta1, meta2;
+                meta1 << "Stage " << sid << " / Feature " << fid;
+                meta2 << "Rectangles: ";
+                Mat temp_window = visualization.clone();
+                Mat temp_metadata = metadata.clone();
+                int current_feature_index = stage_features[sid][fid];
+                current_rects = feature_data[current_feature_index];
+                Mat single_feature = reference_image.clone();
+                resize(single_feature, single_feature, Size(), resize_storage_factor, resize_storage_factor, INTER_LINEAR_EXACT);
+                for(int i = 0; i < (int)current_rects.size(); i++){
+                    rect_data local = current_rects[i];
+                    if(draw_planes){
+                        if(local.weight >= 0){
+                            rectangle(single_feature, Rect(local.x * resize_storage_factor, local.y * resize_storage_factor, local.w * resize_storage_factor, local.h * resize_storage_factor), Scalar(0), FILLED);
+                        }else{
+                            rectangle(single_feature, Rect(local.x * resize_storage_factor, local.y * resize_storage_factor, local.w * resize_storage_factor, local.h * resize_storage_factor), Scalar(255), FILLED);
+                        }
+                    }
+                    Rect part(local.x * resize_factor, local.y * resize_factor, local.w * resize_factor, local.h * resize_factor);
+                    meta2 << part << " (w " << local.weight << ") ";
+                    if(local.weight >= 0){
+                        rectangle(temp_window, part, Scalar(0), FILLED);
+                    }else{
+                        rectangle(temp_window, part, Scalar(255), FILLED);
+                    }
+                }
+                imshow("features", temp_window);
+                putText(temp_window, meta1.str(), Point(15,15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(255));
+                result_video.write(temp_window);
+                // Copy the feature image if needed
+                if(draw_planes){
+                    single_feature.copyTo(image_plane(Rect(0 + (fid%cols_prefered)*single_feature.cols, 0 + (fid/cols_prefered) * single_feature.rows, single_feature.cols, single_feature.rows)));
+                }
+                putText(temp_metadata, meta1.str(), Point(15,15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(255));
+                putText(temp_metadata, meta2.str(), Point(15,40), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(255));
+                imshow("metadata", temp_metadata);
+                waitKey(timing);
+            }
+            //Store the stage image if needed
+            if(draw_planes){
+                stringstream save_location;
+                save_location << output_folder << "stage_" << sid << ".png";
+                imwrite(save_location.str(), image_plane);
+            }
+        }
+    }
+
+    if(lbp){
+        // Grab the corresponding features dimensions and weights
+        FileNode features = cascade["features"];
+        vector<Rect> feature_data;
+        FileNodeIterator it_features = features.begin(), it_features_end = features.end();
+        for(int idf = 0; it_features != it_features_end; it_features++, idf++ ){
+            FileNode rectangle = (*it_features)["rect"];
+            Rect current_feature ((int)rectangle[0], (int)rectangle[1], (int)rectangle[2], (int)rectangle[3]);
+            feature_data.push_back(current_feature);
+        }
+
+        // Loop over each possible feature on its index, visualise on the mask and wait a bit,
+        // then continue to the next feature.
+        Mat image_plane;
+        Mat metadata = Mat::zeros(150, 1000, CV_8UC1);
+        for(int sid = 0; sid < (int)stage_features.size(); sid ++){
+            if(draw_planes){
+                int features_nmbr = (int)stage_features[sid].size();
+                int cols = cols_prefered;
+                int rows = features_nmbr / cols;
+                if( (features_nmbr % cols) > 0){
+                    rows++;
+                }
+                image_plane = Mat::zeros(reference_image.rows * resize_storage_factor * rows, reference_image.cols * resize_storage_factor * cols, CV_8UC1);
+            }
+            for(int fid = 0; fid < (int)stage_features[sid].size(); fid++){
+                stringstream meta1, meta2;
+                meta1 << "Stage " << sid << " / Feature " << fid;
+                meta2 << "Rectangle: ";
+                Mat temp_window = visualization.clone();
+                Mat temp_metadata = metadata.clone();
+                int current_feature_index = stage_features[sid][fid];
+                Rect current_rect = feature_data[current_feature_index];
+                Mat single_feature = reference_image.clone();
+                resize(single_feature, single_feature, Size(), resize_storage_factor, resize_storage_factor, INTER_LINEAR_EXACT);
+
+                // VISUALISATION
+                // The rectangle is the top left one of a 3x3 block LBP constructor
+                Rect resized(current_rect.x * resize_factor, current_rect.y * resize_factor, current_rect.width * resize_factor, current_rect.height * resize_factor);
+                meta2 << resized;
+                // Top left
+                rectangle(temp_window, resized, Scalar(255), 1);
+                // Top middle
+                rectangle(temp_window, Rect(resized.x + resized.width, resized.y, resized.width, resized.height), Scalar(255), 1);
+                // Top right
+                rectangle(temp_window, Rect(resized.x + 2*resized.width, resized.y, resized.width, resized.height), Scalar(255), 1);
+                // Middle left
+                rectangle(temp_window, Rect(resized.x, resized.y + resized.height, resized.width, resized.height), Scalar(255), 1);
+                // Middle middle
+                rectangle(temp_window, Rect(resized.x + resized.width, resized.y + resized.height, resized.width, resized.height), Scalar(255), FILLED);
+                // Middle right
+                rectangle(temp_window, Rect(resized.x + 2*resized.width, resized.y + resized.height, resized.width, resized.height), Scalar(255), 1);
+                // Bottom left
+                rectangle(temp_window, Rect(resized.x, resized.y + 2*resized.height, resized.width, resized.height), Scalar(255), 1);
+                // Bottom middle
+                rectangle(temp_window, Rect(resized.x + resized.width, resized.y + 2*resized.height, resized.width, resized.height), Scalar(255), 1);
+                // Bottom right
+                rectangle(temp_window, Rect(resized.x + 2*resized.width, resized.y + 2*resized.height, resized.width, resized.height), Scalar(255), 1);
+
+                if(draw_planes){
+                    Rect resized_inner(current_rect.x * resize_storage_factor, current_rect.y * resize_storage_factor, current_rect.width * resize_storage_factor, current_rect.height * resize_storage_factor);
+                    // Top left
+                    rectangle(single_feature, resized_inner, Scalar(255), 1);
+                    // Top middle
+                    rectangle(single_feature, Rect(resized_inner.x + resized_inner.width, resized_inner.y, resized_inner.width, resized_inner.height), Scalar(255), 1);
+                    // Top right
+                    rectangle(single_feature, Rect(resized_inner.x + 2*resized_inner.width, resized_inner.y, resized_inner.width, resized_inner.height), Scalar(255), 1);
+                    // Middle left
+                    rectangle(single_feature, Rect(resized_inner.x, resized_inner.y + resized_inner.height, resized_inner.width, resized_inner.height), Scalar(255), 1);
+                    // Middle middle
+                    rectangle(single_feature, Rect(resized_inner.x + resized_inner.width, resized_inner.y + resized_inner.height, resized_inner.width, resized_inner.height), Scalar(255), FILLED);
+                    // Middle right
+                    rectangle(single_feature, Rect(resized_inner.x + 2*resized_inner.width, resized_inner.y + resized_inner.height, resized_inner.width, resized_inner.height), Scalar(255), 1);
+                    // Bottom left
+                    rectangle(single_feature, Rect(resized_inner.x, resized_inner.y + 2*resized_inner.height, resized_inner.width, resized_inner.height), Scalar(255), 1);
+                    // Bottom middle
+                    rectangle(single_feature, Rect(resized_inner.x + resized_inner.width, resized_inner.y + 2*resized_inner.height, resized_inner.width, resized_inner.height), Scalar(255), 1);
+                    // Bottom right
+                    rectangle(single_feature, Rect(resized_inner.x + 2*resized_inner.width, resized_inner.y + 2*resized_inner.height, resized_inner.width, resized_inner.height), Scalar(255), 1);
+
+                    single_feature.copyTo(image_plane(Rect(0 + (fid%cols_prefered)*single_feature.cols, 0 + (fid/cols_prefered) * single_feature.rows, single_feature.cols, single_feature.rows)));
+                }
+
+                putText(temp_metadata, meta1.str(), Point(15,15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(255));
+                putText(temp_metadata, meta2.str(), Point(15,40), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(255));
+                imshow("metadata", temp_metadata);
+                imshow("features", temp_window);
+                putText(temp_window, meta1.str(), Point(15,15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(255));
+                result_video.write(temp_window);
+
+                waitKey(timing);
+            }
+
+            //Store the stage image if needed
+            if(draw_planes){
+                stringstream save_location;
+                save_location << output_folder << "stage_" << sid << ".png";
+                imwrite(save_location.str(), image_plane);
+            }
+        }
+    }
+    return 0;
+}

modules/xobjdetect/tools/waldboost_detector/CMakeLists.txt

@@ -0,0 +1,34 @@
+set(name waldboost_detector)
+set(the_target opencv_${name})
+
+set(OPENCV_${the_target}_DEPS opencv_core opencv_imgproc opencv_imgcodecs opencv_videoio
+    opencv_highgui opencv_xobjdetect)
+
+ocv_check_dependencies(${OPENCV_${the_target}_DEPS})
+
+if(NOT OCV_DEPENDENCIES_FOUND)
+  return()
+endif()
+
+project(${the_target})
+
+ocv_include_directories("${OpenCV_SOURCE_DIR}/include/opencv")
+ocv_include_modules_recurse(${OPENCV_${the_target}_DEPS})
+
+file(GLOB ${the_target}_SOURCES ${CMAKE_CURRENT_SOURCE_DIR}/*.cpp)
+
+add_executable(${the_target} ${${the_target}_SOURCES})
+
+ocv_target_link_libraries(${the_target} ${OPENCV_${the_target}_DEPS})
+
+set_target_properties(${the_target} PROPERTIES
+                      DEBUG_POSTFIX "${OPENCV_DEBUG_POSTFIX}"
+                      ARCHIVE_OUTPUT_DIRECTORY ${LIBRARY_OUTPUT_PATH}
+                      RUNTIME_OUTPUT_DIRECTORY ${EXECUTABLE_OUTPUT_PATH}
+                      OUTPUT_NAME ${the_target})
+
+if(ENABLE_SOLUTION_FOLDERS)
+  set_target_properties(${the_target} PROPERTIES FOLDER "applications")
+endif()
+
+install(TARGETS ${the_target} OPTIONAL RUNTIME DESTINATION bin COMPONENT main)

modules/xobjdetect/tools/waldboost_detector/waldboost_detector.cpp

@@ -0,0 +1,46 @@
+#include "opencv2/xobjdetect.hpp"
+#include "opencv2/imgcodecs.hpp"
+#include "opencv2/imgproc.hpp"
+#include <iostream>
+#include <cstdio>
+using namespace std;
+using namespace cv;
+using namespace cv::xobjdetect;
+
+int main(int argc, char **argv)
+{
+    if (argc < 5) {
+        cerr << "Usage: " << argv[0] << " train <model_filename> <pos_path> <neg_path>" << endl;
+        cerr << "       " << argv[0] << " detect <model_filename> <img_filename> <out_filename> <labelling_filename>" << endl;
+        return 0;
+    }
+
+    string mode = argv[1];
+    Ptr<WBDetector> detector = WBDetector::create();
+    if (mode == "train") {
+        assert(argc == 5);
+        detector->train(argv[3], argv[4]);
+        FileStorage fs(argv[2], FileStorage::WRITE);
+        fs << "waldboost";
+        detector->write(fs);
+    } else if (mode == "detect") {
+        assert(argc == 6);
+        vector<Rect> bboxes;
+        vector<double> confidences;
+        Mat img = imread(argv[3], IMREAD_GRAYSCALE);
+        FileStorage fs(argv[2], FileStorage::READ);
+        detector->read(fs.getFirstTopLevelNode());
+        detector->detect(img, bboxes, confidences);
+
+        FILE *fhandle = fopen(argv[5], "a");
+        for (size_t i = 0; i < bboxes.size(); ++i) {
+            Rect o = bboxes[i];
+            fprintf(fhandle, "%s;%u;%u;%u;%u;%lf\n",
+                argv[3], o.x, o.y, o.width, o.height, confidences[i]);
+        }
+        for (size_t i = 0; i < bboxes.size(); ++i) {
+            rectangle(img, bboxes[i], Scalar(255, 0, 0));
+        }
+        imwrite(argv[4], img);
+    }
+}