renamed OpenCL kernel filename; made some final changes

modules/objdetect/src/cascadedetect.cpp

@@ -1133,7 +1133,7 @@ bool CascadeClassifierImpl::detectSingleScale( InputArray _image, Size processin
 bool CascadeClassifierImpl::ocl_detectSingleScale( InputArray _image, Size processingRectSize,
                                                    int yStep, double factor, Size sumSize0 )
 {
-    const int VECTOR_SIZE = 4;
+    const int VECTOR_SIZE = 1;
     Ptr<HaarEvaluator> haar = featureEvaluator.dynamicCast<HaarEvaluator>();
     if( haar.empty() )
         return false;
@@ -1142,7 +1142,7 @@ bool CascadeClassifierImpl::ocl_detectSingleScale( InputArray _image, Size proce
     
     if( cascadeKernel.empty() )
     {
-        cascadeKernel.create("runHaarClassifierStump", ocl::objdetect::haarobjectdetect_oclsrc,
+        cascadeKernel.create("runHaarClassifierStump", ocl::objdetect::cascadedetect_oclsrc,
                              format("-D VECTOR_SIZE=%d", VECTOR_SIZE));
         if( cascadeKernel.empty() )
             return false;

modules/objdetect/src/opencl/haarobjectdetect.cl → modules/objdetect/src/opencl/cascadedetect.cl

@@ -40,7 +40,7 @@ __kernel void runHaarClassifierStump(
     int2 imgsize, int xyscale, float factor,
     int4 normrect, int2 windowsize, int maxFaces)
 {
-    int ix = get_global_id(0)*xyscale*VECTOR_SIZE;
+    int ix = get_global_id(0)*xyscale;
     int iy = get_global_id(1)*xyscale;
     sumstep /= sizeof(int);
     sqsumstep /= sizeof(int);
@@ -94,7 +94,6 @@ __kernel void runHaarClassifierStump(
         if( stageIdx == nstages )
         {
             int nfaces = atomic_inc(facepos);
-            //printf("detected face #d!!!!\n", nfaces);
             if( nfaces < maxFaces )
             {
                 volatile __global int* face = facepos + 1 + nfaces*4;

samples/cpp/ufacedetect.cpp

@@ -201,13 +201,12 @@ void detectAndDraw( UMat& img, Mat& canvas, CascadeClassifier& cascade,
 
     t = (double)getTickCount();
 
-    cvtColor( img, gray, COLOR_BGR2GRAY );
-    resize( gray, smallImg, Size(), scale0, scale0, INTER_LINEAR );
-    cvtColor(smallImg, canvas, COLOR_GRAY2BGR);
-    equalizeHist( smallImg, smallImg );
+    resize( img, smallImg, Size(), scale0, scale0, INTER_LINEAR );
+    cvtColor( smallImg, gray, COLOR_BGR2GRAY );
+    equalizeHist( gray, gray );
 
-    cascade.detectMultiScale( smallImg, faces,
-        1.1, 2, 0
+    cascade.detectMultiScale( gray, faces,
+        1.1, 3, 0
         //|CASCADE_FIND_BIGGEST_OBJECT
         //|CASCADE_DO_ROUGH_SEARCH
         |CASCADE_SCALE_IMAGE
@@ -215,8 +214,8 @@ void detectAndDraw( UMat& img, Mat& canvas, CascadeClassifier& cascade,
         Size(30, 30) );
     if( tryflip )
     {
-        flip(smallImg, smallImg, 1);
-        cascade.detectMultiScale( smallImg, faces2,
+        flip(gray, gray, 1);
+        cascade.detectMultiScale( gray, faces2,
                                  1.1, 2, 0
                                  //|CASCADE_FIND_BIGGEST_OBJECT
                                  //|CASCADE_DO_ROUGH_SEARCH
@@ -229,7 +228,7 @@ void detectAndDraw( UMat& img, Mat& canvas, CascadeClassifier& cascade,
         }
     }
     t = (double)getTickCount() - t;
-    cvtColor(smallImg, canvas, COLOR_GRAY2BGR);
+    smallImg.copyTo(canvas);
 
     double fps = getTickFrequency()/t;
 
@@ -257,7 +256,7 @@ void detectAndDraw( UMat& img, Mat& canvas, CascadeClassifier& cascade,
                        color, 3, 8, 0);
         if( nestedCascade.empty() )
             continue;
-        UMat smallImgROI = smallImg(*r);
+        UMat smallImgROI = gray(*r);
         nestedCascade.detectMultiScale( smallImgROI, nestedObjects,
             1.1, 2, 0
             //|CASCADE_FIND_BIGGEST_OBJECT