Merge pull request #918 from bitwangyaoyao:2.4_samples

samples/ocl/facedetect.cpp

@@ -1,5 +1,3 @@
-//This sample is inherited from facedetect.cpp in smaple/c
-
 #include "opencv2/objdetect/objdetect.hpp"
 #include "opencv2/highgui/highgui.hpp"
 #include "opencv2/imgproc/imgproc.hpp"
@@ -9,78 +7,84 @@
 
 using namespace std;
 using namespace cv;
+#define LOOP_NUM 10 
 
-static void help()
+const static Scalar colors[] =  { CV_RGB(0,0,255),
-{
+        CV_RGB(0,128,255),
-    cout << "\nThis program demonstrates the cascade recognizer.\n"
+        CV_RGB(0,255,255),
-        "This classifier can recognize many ~rigid objects, it's most known use is for faces.\n"
+        CV_RGB(0,255,0),
-        "Usage:\n"
+        CV_RGB(255,128,0),
-        "./facedetect [--cascade=<cascade_path> this is the primary trained classifier such as frontal face]\n"
+        CV_RGB(255,255,0),
-        "   [--scale=<image scale greater or equal to 1, try 1.3 for example>\n"
+        CV_RGB(255,0,0),
-        "   [filename|camera_index]\n\n"
+        CV_RGB(255,0,255)} ;
-        "see facedetect.cmd for one call:\n"
+
-        "./facedetect --cascade=\"../../data/haarcascades/haarcascade_frontalface_alt.xml\" --scale=1.3 \n"
+int64 work_begin = 0;
-        "Hit any key to quit.\n"
+int64 work_end = 0;
-        "Using OpenCV version " << CV_VERSION << "\n" << endl;
+
+static void workBegin() 
+{ 
+    work_begin = getTickCount();
+}
+static void workEnd()
+{
+    work_end += (getTickCount() - work_begin);
+}
+static double getTime(){
+    return work_end /((double)cvGetTickFrequency() * 1000.);
 }
-struct getRect { Rect operator ()(const CvAvgComp& e) const { return e.rect; } };
-void detectAndDraw( Mat& img,
-    cv::ocl::OclCascadeClassifier& cascade, CascadeClassifier& nestedCascade,
-    double scale);
 
-String cascadeName = "../../../data/haarcascades/haarcascade_frontalface_alt.xml";
+void detect( Mat& img, vector<Rect>& faces, 
+    cv::ocl::OclCascadeClassifierBuf& cascade, 
+    double scale, bool calTime);
+
+void detectCPU( Mat& img, vector<Rect>& faces, 
+    CascadeClassifier& cascade, 
+    double scale, bool calTime);
+
+void Draw(Mat& img, vector<Rect>& faces, double scale);
+
+// This function test if gpu_rst matches cpu_rst.
+// If the two vectors are not equal, it will return the difference in vector size
+// Else if will return (total diff of each cpu and gpu rects covered pixels)/(total cpu rects covered pixels)
+double checkRectSimilarity(Size sz, std::vector<Rect>& cpu_rst, std::vector<Rect>& gpu_rst);
 
 int main( int argc, const char** argv )
 {
+    const char* keys =
+        "{ h | help       | false       | print help message }"
+        "{ i | input      |             | specify input image }"
+        "{ t | template   | ../../../data/haarcascades/haarcascade_frontalface_alt.xml  | specify template file }"
+        "{ c | scale      |   1.0       | scale image }"
+        "{ s | use_cpu    | false       | use cpu or gpu to process the image }";
+
+    CommandLineParser cmd(argc, argv, keys);
+    if (cmd.get<bool>("help"))
+    {
+        cout << "Avaible options:" << endl;
+        cmd.printParams();
+        return 0;
+    }
     CvCapture* capture = 0;
     Mat frame, frameCopy, image;
-    const String scaleOpt = "--scale=";
-    size_t scaleOptLen = scaleOpt.length();
-    const String cascadeOpt = "--cascade=";
-    size_t cascadeOptLen = cascadeOpt.length();
-    String inputName;
 
-    help();
+    bool useCPU = cmd.get<bool>("s");
-    cv::ocl::OclCascadeClassifier cascade;
+    string inputName = cmd.get<string>("i");
-    CascadeClassifier  nestedCascade;
+    string cascadeName = cmd.get<string>("t");
-    double scale = 1;
+    double scale = cmd.get<double>("c");
+    cv::ocl::OclCascadeClassifierBuf cascade;
+    CascadeClassifier  cpu_cascade;
 
-    for( int i = 1; i < argc; i++ )
+    if( !cascade.load( cascadeName ) || !cpu_cascade.load(cascadeName) )
-    {
-        cout << "Processing " << i << " " <<  argv[i] << endl;
-        if( cascadeOpt.compare( 0, cascadeOptLen, argv[i], cascadeOptLen ) == 0 )
-        {
-            cascadeName.assign( argv[i] + cascadeOptLen );
-            cout << "  from which we have cascadeName= " << cascadeName << endl;
-        }
-        else if( scaleOpt.compare( 0, scaleOptLen, argv[i], scaleOptLen ) == 0 )
-        {
-            if( !sscanf( argv[i] + scaleOpt.length(), "%lf", &scale ) || scale < 1 )
-                scale = 1;
-            cout << " from which we read scale = " << scale << endl;
-        }
-        else if( argv[i][0] == '-' )
-        {
-            cerr << "WARNING: Unknown option %s" << argv[i] << endl;
-        }
-        else
-            inputName.assign( argv[i] );
-    }
-
-    if( !cascade.load( cascadeName ) )
     {
         cerr << "ERROR: Could not load classifier cascade" << endl;
-        cerr << "Usage: facedetect [--cascade=<cascade_path>]\n"
-            "   [--scale[=<image scale>\n"
-            "   [filename|camera_index]\n" << endl ;
         return -1;
     }
 
-    if( inputName.empty() || (isdigit(inputName.c_str()[0]) && inputName.c_str()[1] == '\0') )
+    if( inputName.empty() )
     {
-        capture = cvCaptureFromCAM( inputName.empty() ? 0 : inputName.c_str()[0] - '0' );
+        capture = cvCaptureFromCAM(0);
-        int c = inputName.empty() ? 0 : inputName.c_str()[0] - '0' ;
+        if(!capture) 
-        if(!capture) cout << "Capture from CAM " <<  c << " didn't work" << endl;
+            cout << "Capture from CAM 0 didn't work" << endl;
     }
     else if( inputName.size() )
     {
@@ -88,26 +92,30 @@ int main( int argc, const char** argv )
         if( image.empty() )
         {
             capture = cvCaptureFromAVI( inputName.c_str() );
-            if(!capture) cout << "Capture from AVI didn't work" << endl;
+            if(!capture) 
+                cout << "Capture from AVI didn't work" << endl;
+            return -1;
         }
     }
     else
     {
         image = imread( "lena.jpg", 1 );
-        if(image.empty()) cout << "Couldn't read lena.jpg" << endl;
+        if(image.empty()) 
+            cout << "Couldn't read lena.jpg" << endl;
+        return -1;
     }
 
     cvNamedWindow( "result", 1 );
     std::vector<cv::ocl::Info> oclinfo;
     int devnums = cv::ocl::getDevice(oclinfo);
-    if(devnums<1)
+    if( devnums < 1 )
     {
         std::cout << "no device found\n";
         return -1;
     }
     //if you want to use undefault device, set it here
     //setDevice(oclinfo[0]);
-    //setBinpath(CLBINPATH);
+    ocl::setBinpath("./");
     if( capture )
     {
         cout << "In capture ..." << endl;
@@ -115,15 +123,20 @@ int main( int argc, const char** argv )
         {
             IplImage* iplImg = cvQueryFrame( capture );
             frame = iplImg;
+            vector<Rect> faces;
             if( frame.empty() )
                 break;
             if( iplImg->origin == IPL_ORIGIN_TL )
                 frame.copyTo( frameCopy );
             else
                 flip( frame, frameCopy, 0 );
-
+            if(useCPU){
-            detectAndDraw( frameCopy, cascade, nestedCascade, scale );
+                detectCPU(frameCopy, faces, cpu_cascade, scale, false);
-
+            }
+            else{
+                detect(frameCopy, faces, cascade, scale, false);     
+            }
+            Draw(frameCopy, faces, scale);
             if( waitKey( 10 ) >= 0 )
                 goto _cleanup_;
         }
@@ -136,42 +149,34 @@ _cleanup_:
     else
     {
         cout << "In image read" << endl;
-        if( !image.empty() )
+        vector<Rect> faces;
+        vector<Rect> ref_rst;
+        double accuracy = 0.;
+        for(int i = 0; i <= LOOP_NUM;i ++) 
         {
-            detectAndDraw( image, cascade, nestedCascade, scale );
+            cout << "loop" << i << endl;
-            waitKey(0);
+            if(useCPU){
-        }
+                detectCPU(image, faces, cpu_cascade, scale, i==0?false:true);  
-        else if( !inputName.empty() )
+            }
-        {
+            else{
-            /* assume it is a text file containing the
+                detect(image, faces, cascade, scale, i==0?false:true);
-            list of the image filenames to be processed - one per line */
+                if(i == 0){
-            FILE* f = fopen( inputName.c_str(), "rt" );
+                    detectCPU(image, ref_rst, cpu_cascade, scale, false);
-            if( f )
+                    accuracy = checkRectSimilarity(image.size(), ref_rst, faces);
+                }                    
+            }
+            if (i == LOOP_NUM)
             {
-                char buf[1000+1];
+                if (useCPU)
-                while( fgets( buf, 1000, f ) )
+                    cout << "average CPU time (noCamera) : ";
-                {
+                else
-                    int len = (int)strlen(buf), c;
+                    cout << "average GPU time (noCamera) : ";
-                    while( len > 0 && isspace(buf[len-1]) )
+                cout << getTime() / LOOP_NUM << " ms" << endl;
-                        len--;
+                cout << "accuracy value: " << accuracy <<endl;
-                    buf[len] = '\0';
-                    cout << "file " << buf << endl;
-                    image = imread( buf, 1 );
-                    if( !image.empty() )
-                    {
-                        detectAndDraw( image, cascade, nestedCascade, scale );
-                        c = waitKey(0);
-                        if( c == 27 || c == 'q' || c == 'Q' )
-                            break;
-                    }
-                    else
-                    {
-                        cerr << "Aw snap, couldn't read image " << buf << endl;
-                    }
-                }
-                fclose(f);
             }
         }
+        Draw(image, faces, scale);
+        waitKey(0);
     }
 
     cvDestroyWindow("result");
@@ -179,44 +184,44 @@ _cleanup_:
     return 0;
 }
 
-void detectAndDraw( Mat& img,
+void detect( Mat& img, vector<Rect>& faces, 
-    cv::ocl::OclCascadeClassifier& cascade, CascadeClassifier&,
+    cv::ocl::OclCascadeClassifierBuf& cascade, 
-    double scale)
+    double scale, bool calTime)
 {
-    int i = 0;
-    double t = 0;
-    vector<Rect> faces;
-    const static Scalar colors[] =  { CV_RGB(0,0,255),
-        CV_RGB(0,128,255),
-        CV_RGB(0,255,255),
-        CV_RGB(0,255,0),
-        CV_RGB(255,128,0),
-        CV_RGB(255,255,0),
-        CV_RGB(255,0,0),
-        CV_RGB(255,0,255)} ;
     cv::ocl::oclMat image(img);
     cv::ocl::oclMat gray, smallImg( cvRound (img.rows/scale), cvRound(img.cols/scale), CV_8UC1 );
-
+    if(calTime) workBegin();
     cv::ocl::cvtColor( image, gray, CV_BGR2GRAY );
     cv::ocl::resize( gray, smallImg, smallImg.size(), 0, 0, INTER_LINEAR );
     cv::ocl::equalizeHist( smallImg, smallImg );
 
-    CvSeq* _objects;
+    cascade.detectMultiScale( smallImg, faces, 1.1,
-    MemStorage storage(cvCreateMemStorage(0));
-    t = (double)cvGetTickCount();
-    _objects = cascade.oclHaarDetectObjects( smallImg, storage, 1.1,
         3, 0
         |CV_HAAR_SCALE_IMAGE
         , Size(30,30), Size(0, 0) );
-    vector<CvAvgComp> vecAvgComp;
+    if(calTime) workEnd();
-    Seq<CvAvgComp>(_objects).copyTo(vecAvgComp);
+}
-    faces.resize(vecAvgComp.size());
+
-    std::transform(vecAvgComp.begin(), vecAvgComp.end(), faces.begin(), getRect());
+void detectCPU( Mat& img, vector<Rect>& faces, 
-    t = (double)cvGetTickCount() - t;
+    CascadeClassifier& cascade, 
-    printf( "detection time = %g ms\n", t/((double)cvGetTickFrequency()*1000.) );
+    double scale, bool calTime)
+{
+    if(calTime) workBegin();
+    Mat cpu_gray, cpu_smallImg( cvRound (img.rows/scale), cvRound(img.cols/scale), CV_8UC1 );
+    cvtColor(img, cpu_gray, CV_BGR2GRAY);
+    resize(cpu_gray, cpu_smallImg, cpu_smallImg.size(), 0, 0, INTER_LINEAR);
+    equalizeHist(cpu_smallImg, cpu_smallImg);
+    cascade.detectMultiScale(cpu_smallImg, faces, 1.1,
+        3, 0 | CV_HAAR_SCALE_IMAGE,
+        Size(30, 30), Size(0, 0));
+    if(calTime) workEnd(); 
+}
+
+void Draw(Mat& img, vector<Rect>& faces, double scale)
+{
+    int i = 0;
     for( vector<Rect>::const_iterator r = faces.begin(); r != faces.end(); r++, i++ )
     {
-        Mat smallImgROI;
         Point center;
         Scalar color = colors[i%8];
         int radius;
@@ -227,3 +232,42 @@ void detectAndDraw( Mat& img,
     }
     cv::imshow( "result", img );
 }
+
+double checkRectSimilarity(Size sz, std::vector<Rect>& ob1, std::vector<Rect>& ob2)
+{
+    double final_test_result = 0.0;
+    size_t sz1 = ob1.size();
+    size_t sz2 = ob2.size();
+
+    if(sz1 != sz2)
+        return sz1 > sz2 ? (double)(sz1 - sz2) : (double)(sz2 - sz1);
+    else
+    {
+        cv::Mat cpu_result(sz, CV_8UC1);
+        cpu_result.setTo(0);
+
+        for(vector<Rect>::const_iterator r = ob1.begin(); r != ob1.end(); r++)
+        {      
+            cv::Mat cpu_result_roi(cpu_result, *r);
+            cpu_result_roi.setTo(1);
+            cpu_result.copyTo(cpu_result);
+        }
+        int cpu_area = cv::countNonZero(cpu_result > 0);
+
+        cv::Mat gpu_result(sz, CV_8UC1);
+        gpu_result.setTo(0);
+        for(vector<Rect>::const_iterator r2 = ob2.begin(); r2 != ob2.end(); r2++)
+        {
+            cv::Mat gpu_result_roi(gpu_result, *r2);
+            gpu_result_roi.setTo(1);
+            gpu_result.copyTo(gpu_result);
+        }
+
+        cv::Mat result_;
+        multiply(cpu_result, gpu_result, result_);
+        int result = cv::countNonZero(result_ > 0);
+
+        final_test_result = 1.0 - (double)result/(double)cpu_area;
+    }
+    return final_test_result;
+}

samples/ocl/hog.cpp

@@ -45,7 +45,6 @@ public:
     bool gamma_corr;
 };
 
-
 class App
 {
 public:
@@ -64,6 +63,13 @@ public:
 
     string message() const;
 
+// This function test if gpu_rst matches cpu_rst.
+// If the two vectors are not equal, it will return the difference in vector size
+// Else if will return 
+// (total diff of each cpu and gpu rects covered pixels)/(total cpu rects covered pixels)
+    double checkRectSimilarity(Size sz, 
+                               std::vector<Rect>& cpu_rst, 
+                               std::vector<Rect>& gpu_rst);
 private:
     App operator=(App&);
 
@@ -290,6 +296,7 @@ void App::run()
         ocl::oclMat gpu_img;
 
         // Iterate over all frames
+        bool verify = false;
         while (running && !frame.empty())
         {
             workBegin();
@@ -316,7 +323,18 @@ void App::run()
                 gpu_img.upload(img);
                 gpu_hog.detectMultiScale(gpu_img, found, hit_threshold, win_stride,
                                          Size(0, 0), scale, gr_threshold);
-            }
+                if (!verify)
+                {
+                    // verify if GPU output same objects with CPU at 1st run
+                    verify = true;
+                    vector<Rect> ref_rst;
+                    cvtColor(img, img, CV_BGRA2BGR);
+                    cpu_hog.detectMultiScale(img, ref_rst, hit_threshold, win_stride,
+                                              Size(0, 0), scale, gr_threshold-2);
+                    double accuracy = checkRectSimilarity(img.size(), ref_rst, found);
+                    cout << "\naccuracy value: " << accuracy << endl;           
+                } 
+           }
             else cpu_hog.detectMultiScale(img, found, hit_threshold, win_stride,
                                           Size(0, 0), scale, gr_threshold);
             hogWorkEnd();
@@ -457,3 +475,45 @@ inline string App::workFps() const
     return ss.str();
 }
 
+double App::checkRectSimilarity(Size sz, 
+                                std::vector<Rect>& ob1, 
+                                std::vector<Rect>& ob2)
+{
+    double final_test_result = 0.0;
+    size_t sz1 = ob1.size();
+    size_t sz2 = ob2.size();
+
+    if(sz1 != sz2)
+        return sz1 > sz2 ? (double)(sz1 - sz2) : (double)(sz2 - sz1);
+    else
+    {
+        cv::Mat cpu_result(sz, CV_8UC1);
+        cpu_result.setTo(0);
+
+        for(vector<Rect>::const_iterator r = ob1.begin(); r != ob1.end(); r++)
+        {      
+            cv::Mat cpu_result_roi(cpu_result, *r);
+            cpu_result_roi.setTo(1);
+            cpu_result.copyTo(cpu_result);
+        }
+        int cpu_area = cv::countNonZero(cpu_result > 0);
+
+        cv::Mat gpu_result(sz, CV_8UC1);
+        gpu_result.setTo(0);
+        for(vector<Rect>::const_iterator r2 = ob2.begin(); r2 != ob2.end(); r2++)
+        {
+            cv::Mat gpu_result_roi(gpu_result, *r2);
+            gpu_result_roi.setTo(1);
+            gpu_result.copyTo(gpu_result);
+        }
+
+        cv::Mat result_;
+        multiply(cpu_result, gpu_result, result_);
+        int result = cv::countNonZero(result_ > 0);
+
+        final_test_result = 1.0 - (double)result/(double)cpu_area;
+    }
+    return final_test_result;
+
+}
+

samples/ocl/pyrlk_optical_flow.cpp

@@ -0,0 +1,290 @@
+#include <iostream>
+#include <vector>
+#include <iomanip>
+
+#include "opencv2/highgui/highgui.hpp"
+#include "opencv2/ocl/ocl.hpp"
+#include "opencv2/video/video.hpp"
+
+using namespace std;
+using namespace cv;
+using namespace cv::ocl;
+
+typedef unsigned char uchar;
+#define LOOP_NUM 10 
+int64 work_begin = 0;
+int64 work_end = 0;
+
+static void workBegin() 
+{ 
+    work_begin = getTickCount();
+}
+static void workEnd()
+{
+    work_end += (getTickCount() - work_begin);
+}
+static double getTime(){
+    return work_end * 1000. / getTickFrequency();
+}
+
+static void download(const oclMat& d_mat, vector<Point2f>& vec)
+{
+    vec.resize(d_mat.cols);
+    Mat mat(1, d_mat.cols, CV_32FC2, (void*)&vec[0]);
+    d_mat.download(mat);
+}
+
+static void download(const oclMat& d_mat, vector<uchar>& vec)
+{
+    vec.resize(d_mat.cols);
+    Mat mat(1, d_mat.cols, CV_8UC1, (void*)&vec[0]);
+    d_mat.download(mat);
+}
+
+static void drawArrows(Mat& frame, const vector<Point2f>& prevPts, const vector<Point2f>& nextPts, const vector<uchar>& status, Scalar line_color = Scalar(0, 0, 255))
+{
+    for (size_t i = 0; i < prevPts.size(); ++i)
+    {
+        if (status[i])
+        {
+            int line_thickness = 1;
+
+            Point p = prevPts[i];
+            Point q = nextPts[i];
+
+            double angle = atan2((double) p.y - q.y, (double) p.x - q.x);
+
+            double hypotenuse = sqrt( (double)(p.y - q.y)*(p.y - q.y) + (double)(p.x - q.x)*(p.x - q.x) );
+
+            if (hypotenuse < 1.0)
+                continue;
+
+            // Here we lengthen the arrow by a factor of three.
+            q.x = (int) (p.x - 3 * hypotenuse * cos(angle));
+            q.y = (int) (p.y - 3 * hypotenuse * sin(angle));
+
+            // Now we draw the main line of the arrow.
+            line(frame, p, q, line_color, line_thickness);
+
+            // Now draw the tips of the arrow. I do some scaling so that the
+            // tips look proportional to the main line of the arrow.
+
+            p.x = (int) (q.x + 9 * cos(angle + CV_PI / 4));
+            p.y = (int) (q.y + 9 * sin(angle + CV_PI / 4));
+            line(frame, p, q, line_color, line_thickness);
+
+            p.x = (int) (q.x + 9 * cos(angle - CV_PI / 4));
+            p.y = (int) (q.y + 9 * sin(angle - CV_PI / 4));
+            line(frame, p, q, line_color, line_thickness);
+        }
+    }
+}
+
+
+int main(int argc, const char* argv[])
+{
+    static std::vector<Info> ocl_info;
+    ocl::getDevice(ocl_info);
+    //if you want to use undefault device, set it here
+    setDevice(ocl_info[0]);
+
+    //set this to save kernel compile time from second time you run
+    ocl::setBinpath("./");
+    const char* keys =
+        "{ h            | help           | false | print help message }"
+        "{ l            | left           |       | specify left image }"
+        "{ r            | right          |       | specify right image }"
+        "{ c            | camera         | 0     | enable camera capturing }"
+        "{ s            | use_cpu        | false | use cpu or gpu to process the image }"
+        "{ v            | video          |       | use video as input }"
+        "{ points       | points         | 1000  | specify points count [GoodFeatureToTrack] }"
+        "{ min_dist     | min_dist       | 0     | specify minimal distance between points [GoodFeatureToTrack] }";
+
+    CommandLineParser cmd(argc, argv, keys);
+
+    if (cmd.get<bool>("help"))
+    {
+        cout << "Usage: pyrlk_optical_flow [options]" << endl;
+        cout << "Avaible options:" << endl;
+        cmd.printParams();
+        return 0;
+    }
+
+    bool defaultPicturesFail = false;
+    string fname0 = cmd.get<string>("left");
+    string fname1 = cmd.get<string>("right");
+    string vdofile = cmd.get<string>("video");
+    int points = cmd.get<int>("points");
+    double minDist = cmd.get<double>("min_dist");
+    bool useCPU = cmd.get<bool>("s");
+    bool useCamera = cmd.get<bool>("c");
+    int inputName = cmd.get<int>("c");
+    oclMat d_nextPts, d_status;
+
+    Mat frame0 = imread(fname0, cv::IMREAD_GRAYSCALE);
+    Mat frame1 = imread(fname1, cv::IMREAD_GRAYSCALE);
+    PyrLKOpticalFlow d_pyrLK;
+    vector<cv::Point2f> pts;
+    vector<cv::Point2f> nextPts;
+    vector<unsigned char> status;
+    vector<float> err;
+
+    if (frame0.empty() || frame1.empty())
+    {
+        useCamera = true;
+        defaultPicturesFail = true;
+        CvCapture* capture = 0;
+        capture = cvCaptureFromCAM( inputName );
+        if (!capture)
+        {
+            cout << "Can't load input images" << endl;
+            return -1;
+        }
+    }
+
+    cout << "Points count : " << points << endl << endl;
+
+    if (useCamera)
+    {
+        CvCapture* capture = 0;
+        Mat frame, frameCopy;
+        Mat frame0Gray, frame1Gray;
+        Mat ptr0, ptr1;
+
+        if(vdofile == "")
+            capture = cvCaptureFromCAM( inputName );
+        else
+            capture = cvCreateFileCapture(vdofile.c_str());
+
+        int c = inputName ;
+        if(!capture)
+        {
+            if(vdofile == "")
+                cout << "Capture from CAM " << c << " didn't work" << endl;
+            else
+                cout << "Capture from file " << vdofile << " failed" <<endl;
+            if (defaultPicturesFail)
+            {
+                return -1;
+            }
+            goto nocamera;
+        }
+
+        cout << "In capture ..." << endl;
+        for(int i = 0;; i++)
+        {
+            frame = cvQueryFrame( capture );
+            if( frame.empty() )
+                break;
+
+            if (i == 0)
+            {
+                frame.copyTo( frame0 );
+                cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
+            }
+            else
+            {
+                if (i%2 == 1)
+                {
+                    frame.copyTo(frame1);
+                    cvtColor(frame1, frame1Gray, COLOR_BGR2GRAY);
+                    ptr0 = frame0Gray;
+                    ptr1 = frame1Gray;
+                }
+                else
+                {
+                    frame.copyTo(frame0);
+                    cvtColor(frame0, frame0Gray, COLOR_BGR2GRAY);
+                    ptr0 = frame1Gray;
+                    ptr1 = frame0Gray;
+                }
+
+                pts.clear();
+
+                cv::goodFeaturesToTrack(ptr0, pts, points, 0.01, 0.0);
+
+                if (pts.size() == 0)
+                {
+                    continue;
+                }
+
+                if (useCPU)
+                {
+                    cv::calcOpticalFlowPyrLK(ptr0, ptr1, pts, nextPts, status, err);
+                }
+                else
+                {
+                    oclMat d_prevPts(1, points, CV_32FC2, (void*)&pts[0]);
+
+                    d_pyrLK.sparse(oclMat(ptr0), oclMat(ptr1), d_prevPts, d_nextPts, d_status);
+
+                    download(d_prevPts, pts);
+                    download(d_nextPts, nextPts);
+                    download(d_status, status);
+
+                }
+                if (i%2 == 1)
+                    frame1.copyTo(frameCopy);
+                else
+                    frame0.copyTo(frameCopy);
+                drawArrows(frameCopy, pts, nextPts, status, Scalar(255, 0, 0));
+                imshow("PyrLK [Sparse]", frameCopy);
+            }
+
+            if( waitKey( 10 ) >= 0 )
+                goto _cleanup_;
+        }
+
+        waitKey(0);
+
+_cleanup_:
+        cvReleaseCapture( &capture );
+    }
+    else
+    {
+nocamera:
+        for(int i = 0; i <= LOOP_NUM;i ++) 
+        {
+            cout << "loop" << i << endl;
+            if (i > 0) workBegin();
+
+            cv::goodFeaturesToTrack(frame0, pts, points, 0.01, minDist);
+
+            if (useCPU)
+            {
+                cv::calcOpticalFlowPyrLK(frame0, frame1, pts, nextPts, status, err);
+            }
+            else
+            {
+                oclMat d_prevPts(1, points, CV_32FC2, (void*)&pts[0]);
+
+                d_pyrLK.sparse(oclMat(frame0), oclMat(frame1), d_prevPts, d_nextPts, d_status);
+
+                download(d_prevPts, pts);
+                download(d_nextPts, nextPts);
+                download(d_status, status);
+            }
+
+            if (i > 0 && i <= LOOP_NUM)
+                workEnd();
+
+            if (i == LOOP_NUM)
+            {
+                if (useCPU)
+                    cout << "average CPU time (noCamera) : ";
+                else
+                    cout << "average GPU time (noCamera) : ";
+
+                cout << getTime() / LOOP_NUM << " ms" << endl;
+
+                drawArrows(frame0, pts, nextPts, status, Scalar(255, 0, 0));
+
+                imshow("PyrLK [Sparse]", frame0);
+            }
+        }
+    }
+
+    waitKey();
+
+    return 0;
+}

samples/ocl/stereo_match.cpp

@@ -0,0 +1,419 @@
+#include <iostream>
+#include <string>
+#include <sstream>
+#include <iomanip>
+#include <stdexcept>
+#include "opencv2/ocl/ocl.hpp"
+#include "opencv2/highgui/highgui.hpp"
+
+using namespace cv;
+using namespace std;
+using namespace ocl;
+
+bool help_showed = false;
+
+struct Params
+{
+    Params();
+    static Params read(int argc, char** argv);
+
+    string left;
+    string right;
+
+    string method_str() const
+    {
+        switch (method)
+        {
+        case BM: return "BM";
+        case BP: return "BP";
+        case CSBP: return "CSBP";
+        }
+        return "";
+    }
+    enum {BM, BP, CSBP} method;
+    int ndisp; // Max disparity + 1
+    enum {GPU, CPU} type;
+};
+
+
+struct App
+{
+    App(const Params& p);
+    void run();
+    void handleKey(char key);
+    void printParams() const;
+
+    void workBegin() { work_begin = getTickCount(); }
+    void workEnd()
+    {
+        int64 d = getTickCount() - work_begin;
+        double f = getTickFrequency();
+        work_fps = f / d;
+    }
+
+    string text() const
+    {
+        stringstream ss;
+        ss << "(" << p.method_str() << ") FPS: " << setiosflags(ios::left)
+            << setprecision(4) << work_fps;
+        return ss.str();
+    }
+private:
+    Params p;
+    bool running;
+
+    Mat left_src, right_src;
+    Mat left, right;
+    oclMat d_left, d_right;
+
+    StereoBM_OCL bm;
+    StereoBeliefPropagation bp;
+    StereoConstantSpaceBP csbp;
+
+    int64 work_begin;
+    double work_fps;
+};
+
+static void printHelp()
+{
+    cout << "Usage: stereo_match_gpu\n"
+        << "\t--left <left_view> --right <right_view> # must be rectified\n"
+        << "\t--method <stereo_match_method> # BM | BP | CSBP\n"
+        << "\t--ndisp <number> # number of disparity levels\n"
+        << "\t--type <device_type> # cpu | CPU | gpu | GPU\n";
+    help_showed = true;
+}
+
+int main(int argc, char** argv)
+{
+    try
+    {
+        if (argc < 2)
+        {
+            printHelp();
+            return 1;
+        }
+
+        Params args = Params::read(argc, argv);
+        if (help_showed)
+            return -1;
+
+        int flags[2] = { CVCL_DEVICE_TYPE_GPU, CVCL_DEVICE_TYPE_CPU };
+        vector<Info> info;
+
+        if(getDevice(info, flags[args.type]) == 0)
+        {
+            throw runtime_error("Error: Did not find a valid OpenCL device!");
+        }
+        cout << "Device name:" << info[0].DeviceName[0] << endl;
+
+        App app(args);
+        app.run();
+    }
+    catch (const exception& e)
+    {
+        cout << "error: " << e.what() << endl;
+    }
+    return 0;
+}
+
+
+Params::Params()
+{
+    method = BM;
+    ndisp = 64;
+    type = GPU;
+}
+
+
+Params Params::read(int argc, char** argv)
+{
+    Params p;
+
+    for (int i = 1; i < argc; i++)
+    {
+        if (string(argv[i]) == "--left") p.left = argv[++i];
+        else if (string(argv[i]) == "--right") p.right = argv[++i];
+        else if (string(argv[i]) == "--method")
+        {
+            if (string(argv[i + 1]) == "BM") p.method = BM;
+            else if (string(argv[i + 1]) == "BP") p.method = BP;
+            else if (string(argv[i + 1]) == "CSBP") p.method = CSBP;
+            else throw runtime_error("unknown stereo match method: " + string(argv[i + 1]));
+            i++;
+        }
+        else if (string(argv[i]) == "--ndisp") p.ndisp = atoi(argv[++i]);
+        else if (string(argv[i]) == "--type")
+        {
+            string t(argv[++i]);
+            if (t == "cpu" || t == "CPU")
+            {
+                p.type = CPU;
+            } 
+            else if (t == "gpu" || t == "GPU")
+            {
+                p.type = GPU;
+            }
+            else throw runtime_error("unknown device type: " + t);
+        }
+        else if (string(argv[i]) == "--help") printHelp();
+        else throw runtime_error("unknown key: " + string(argv[i]));
+    }
+
+    return p;
+}
+
+
+App::App(const Params& params)
+    : p(params), running(false)
+{
+    cout << "stereo_match_ocl sample\n";
+    cout << "\nControls:\n"
+        << "\tesc - exit\n"
+        << "\tp - print current parameters\n"
+        << "\tg - convert source images into gray\n"
+        << "\tm - change stereo match method\n"
+        << "\ts - change Sobel prefiltering flag (for BM only)\n"
+        << "\t1/q - increase/decrease maximum disparity\n"
+        << "\t2/w - increase/decrease window size (for BM only)\n"
+        << "\t3/e - increase/decrease iteration count (for BP and CSBP only)\n"
+        << "\t4/r - increase/decrease level count (for BP and CSBP only)\n";
+}
+
+
+void App::run()
+{
+    // Load images
+    left_src = imread(p.left);
+    right_src = imread(p.right);
+    if (left_src.empty()) throw runtime_error("can't open file \"" + p.left + "\"");
+    if (right_src.empty()) throw runtime_error("can't open file \"" + p.right + "\"");
+
+    cvtColor(left_src, left, CV_BGR2GRAY);
+    cvtColor(right_src, right, CV_BGR2GRAY);
+
+    d_left.upload(left);
+    d_right.upload(right);
+
+    imshow("left", left);
+    imshow("right", right);
+
+    // Set common parameters
+    bm.ndisp = p.ndisp;
+    bp.ndisp = p.ndisp;
+    csbp.ndisp = p.ndisp;
+
+    cout << endl;
+    printParams();
+
+    running = true;
+    while (running)
+    {
+
+        // Prepare disparity map of specified type
+        Mat disp;
+        oclMat d_disp;
+        workBegin();
+        switch (p.method)
+        {
+        case Params::BM:
+            if (d_left.channels() > 1 || d_right.channels() > 1)
+            {
+                cout << "BM doesn't support color images\n";
+                cvtColor(left_src, left, CV_BGR2GRAY);
+                cvtColor(right_src, right, CV_BGR2GRAY);
+                cout << "image_channels: " << left.channels() << endl;
+                d_left.upload(left);
+                d_right.upload(right);
+                imshow("left", left);
+                imshow("right", right);
+            }
+            bm(d_left, d_right, d_disp);
+            break;
+        case Params::BP:
+            bp(d_left, d_right, d_disp);
+            break;
+        case Params::CSBP:
+            csbp(d_left, d_right, d_disp);
+            break;
+        }
+        ocl::finish();
+        workEnd();
+
+        // Show results
+        d_disp.download(disp);
+        if (p.method != Params::BM)
+        {
+            disp.convertTo(disp, 0);
+        }
+        putText(disp, text(), Point(5, 25), FONT_HERSHEY_SIMPLEX, 1.0, Scalar::all(255));
+        imshow("disparity", disp);
+
+        handleKey((char)waitKey(3));
+    }
+}
+
+
+void App::printParams() const
+{
+    cout << "--- Parameters ---\n";
+    cout << "image_size: (" << left.cols << ", " << left.rows << ")\n";
+    cout << "image_channels: " << left.channels() << endl;
+    cout << "method: " << p.method_str() << endl
+        << "ndisp: " << p.ndisp << endl;
+    switch (p.method)
+    {
+    case Params::BM:
+        cout << "win_size: " << bm.winSize << endl;
+        cout << "prefilter_sobel: " << bm.preset << endl;
+        break;
+    case Params::BP:
+        cout << "iter_count: " << bp.iters << endl;
+        cout << "level_count: " << bp.levels << endl;
+        break;
+    case Params::CSBP:
+        cout << "iter_count: " << csbp.iters << endl;
+        cout << "level_count: " << csbp.levels << endl;
+        break;
+    }
+    cout << endl;
+}
+
+
+void App::handleKey(char key)
+{
+    switch (key)
+    {
+    case 27:
+        running = false;
+        break;
+    case 'p': case 'P':
+        printParams();
+        break;
+    case 'g': case 'G':
+        if (left.channels() == 1 && p.method != Params::BM)
+        {
+            left = left_src;
+            right = right_src;
+        }
+        else
+        {
+            cvtColor(left_src, left, CV_BGR2GRAY);
+            cvtColor(right_src, right, CV_BGR2GRAY);
+        }
+        d_left.upload(left);
+        d_right.upload(right);
+        cout << "image_channels: " << left.channels() << endl;
+        imshow("left", left);
+        imshow("right", right);
+        break;
+    case 'm': case 'M':
+        switch (p.method)
+        {
+        case Params::BM:
+            p.method = Params::BP;
+            break;
+        case Params::BP:
+            p.method = Params::CSBP;
+            break;
+        case Params::CSBP:
+            p.method = Params::BM;
+            break;
+        }
+        cout << "method: " << p.method_str() << endl;
+        break;
+    case 's': case 'S':
+        if (p.method == Params::BM)
+        {
+            switch (bm.preset)
+            {
+            case StereoBM_OCL::BASIC_PRESET:
+                bm.preset = StereoBM_OCL::PREFILTER_XSOBEL;
+                break;
+            case StereoBM_OCL::PREFILTER_XSOBEL:
+                bm.preset = StereoBM_OCL::BASIC_PRESET;
+                break;
+            }
+            cout << "prefilter_sobel: " << bm.preset << endl;
+        }
+        break;
+    case '1':
+        p.ndisp = p.ndisp == 1 ? 8 : p.ndisp + 8;
+        cout << "ndisp: " << p.ndisp << endl;
+        bm.ndisp = p.ndisp;
+        bp.ndisp = p.ndisp;
+        csbp.ndisp = p.ndisp;
+        break;
+    case 'q': case 'Q':
+        p.ndisp = max(p.ndisp - 8, 1);
+        cout << "ndisp: " << p.ndisp << endl;
+        bm.ndisp = p.ndisp;
+        bp.ndisp = p.ndisp;
+        csbp.ndisp = p.ndisp;
+        break;
+    case '2':
+        if (p.method == Params::BM)
+        {
+            bm.winSize = min(bm.winSize + 1, 51);
+            cout << "win_size: " << bm.winSize << endl;
+        }
+        break;
+    case 'w': case 'W':
+        if (p.method == Params::BM)
+        {
+            bm.winSize = max(bm.winSize - 1, 2);
+            cout << "win_size: " << bm.winSize << endl;
+        }
+        break;
+    case '3':
+        if (p.method == Params::BP)
+        {
+            bp.iters += 1;
+            cout << "iter_count: " << bp.iters << endl;
+        }
+        else if (p.method == Params::CSBP)
+        {
+            csbp.iters += 1;
+            cout << "iter_count: " << csbp.iters << endl;
+        }
+        break;
+    case 'e': case 'E':
+        if (p.method == Params::BP)
+        {
+            bp.iters = max(bp.iters - 1, 1);
+            cout << "iter_count: " << bp.iters << endl;
+        }
+        else if (p.method == Params::CSBP)
+        {
+            csbp.iters = max(csbp.iters - 1, 1);
+            cout << "iter_count: " << csbp.iters << endl;
+        }
+        break;
+    case '4':
+        if (p.method == Params::BP)
+        {
+            bp.levels += 1;
+            cout << "level_count: " << bp.levels << endl;
+        }
+        else if (p.method == Params::CSBP)
+        {
+            csbp.levels += 1;
+            cout << "level_count: " << csbp.levels << endl;
+        }
+        break;
+    case 'r': case 'R':
+        if (p.method == Params::BP)
+        {
+            bp.levels = max(bp.levels - 1, 1);
+            cout << "level_count: " << bp.levels << endl;
+        }
+        else if (p.method == Params::CSBP)
+        {
+            csbp.levels = max(csbp.levels - 1, 1);
+            cout << "level_count: " << csbp.levels << endl;
+        }
+        break;
+    }
+}
+
+