generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

continue adding OpenCL optimization to cascade classifier

Browse Source
This commit is contained in:
Vadim Pisarevsky
2013-12-12 21:58:42 +04:00
parent 302a5adcc2
commit d8513d627d
3 changed files with 273 additions and 116 deletions
Download Patch File Download Diff File Expand all files Collapse all files

288
modules/objdetect/src/cascadedetect.cpp
View File

@@ -7,10 +7,10 @@
// copy or use the software.
//
//
// Intel License Agreement
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2000, Intel Corporation, all rights reserved.
// Copyright (C) 2008-2013, 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,
@@ -23,13 +23,13 @@
// 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
// * The name of Itseez Inc. 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,
// In no event shall the 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
@@ -434,7 +434,7 @@ FeatureEvaluator::~FeatureEvaluator() {}
bool FeatureEvaluator::read(const FileNode&) {return true;}
Ptr<FeatureEvaluator> FeatureEvaluator::clone() const { return Ptr<FeatureEvaluator>(); }
int FeatureEvaluator::getFeatureType() const {return -1;}
bool FeatureEvaluator::setImage(const Mat&, Size) {return true;}
bool FeatureEvaluator::setImage(InputArray, Size) {return true;}
bool FeatureEvaluator::setWindow(Point) { return true; }
double FeatureEvaluator::calcOrd(int) const { return 0.; }
int FeatureEvaluator::calcCat(int) const { return 0; }
@@ -466,7 +466,9 @@ bool HaarEvaluator::Feature :: read( const FileNode& node )
HaarEvaluator::HaarEvaluator()
{
features = makePtr<std::vector<Feature> >();
optfeaturesPtr = 0;
pwin = 0;
pqwin = 0;
}
HaarEvaluator::~HaarEvaluator()
{
@@ -476,16 +478,16 @@ bool HaarEvaluator::read(const FileNode& node)
{
size_t i, n = node.size();
CV_Assert(n > 0);
features.resize(n);
featuresPtr = &features[0];
features->resize(n);
FileNodeIterator it = node.begin();
hasTiltedFeatures = false;
std::vector<Feature> ff = *features;
for(i = 0; i < n; i++, ++it)
{
if(!features[i].read(*it))
if(!ff[i].read(*it))
return false;
if( features[i].tilted )
if( ff[i].tilted )
hasTiltedFeatures = true;
}
return true;
@@ -496,55 +498,60 @@ Ptr<FeatureEvaluator> HaarEvaluator::clone() const
Ptr<HaarEvaluator> ret = makePtr<HaarEvaluator>();
ret->origWinSize = origWinSize;
ret->features = features;
ret->optfeatures = optfeatures;
ret->optfeaturesPtr = optfeatures->empty() ? 0 : &(*(ret->optfeatures))[0];
ret->hasTiltedFeatures = hasTiltedFeatures;
ret->sum0 = sum0, ret->sqsum0 = sqsum0, ret->tilted0 = tilted0;
ret->sum = sum, ret->sqsum = sqsum, ret->tilted = tilted;
ret->sum0 = sum0; ret->sqsum0 = sqsum0;
ret->sum = sum; ret->sqsum = sqsum; ret->tilted = tilted;
ret->normrect = normrect;
memcpy( ret->p, p, 4*sizeof(p[0]) );
memcpy( ret->pq, pq, 4*sizeof(pq[0]) );
ret->offset = offset;
memcpy( ret->nofs, nofs, 4*sizeof(nofs[0]) );
memcpy( ret->nqofs, nqofs, 4*sizeof(nqofs[0]) );
ret->pwin = pwin; ret->pqwin = pqwin;
ret->varianceNormFactor = varianceNormFactor;
return ret;
}
bool HaarEvaluator::setImage( const Mat &image, Size _origWinSize )
bool HaarEvaluator::setImage( InputArray _image, Size _origWinSize )
{
int rn = image.rows+1, cn = image.cols+1;
Size imgsz = _image.size();
int rn = imgsz.height+1, cn = imgsz.width+1, rnt = rn;
origWinSize = _origWinSize;
normrect = Rect(1, 1, origWinSize.width-2, origWinSize.height-2);
if (image.cols < origWinSize.width || image.rows < origWinSize.height)
if (imgsz.width < origWinSize.width || imgsz.height < origWinSize.height)
return false;
if( sum0.rows < rn || sum0.cols < cn )
if( hasTiltedFeatures )
rnt = rn*2;
if( sum0.rows < rnt || sum0.cols < cn )
{
sum0.create(rn, cn, CV_32S);
sum0.create(rnt, cn, CV_32S);
sqsum0.create(rn, cn, CV_64F);
if (hasTiltedFeatures)
tilted0.create( rn, cn, CV_32S);
}
sum = Mat(rn, cn, CV_32S, sum0.data);
sqsum = Mat(rn, cn, CV_64F, sqsum0.data);
if( hasTiltedFeatures )
{
tilted = Mat(rn, cn, CV_32S, tilted0.data);
integral(image, sum, sqsum, tilted);
tilted = Mat(rn, cn, CV_32S, sum0.data + rn*sum.step);
integral(_image, sum, sqsum, tilted);
}
else
integral(image, sum, sqsum);
const int* sdata = (const int*)sum.data;
const double* sqdata = (const double*)sqsum.data;
size_t sumStep = sum.step/sizeof(sdata[0]);
size_t sqsumStep = sqsum.step/sizeof(sqdata[0]);
integral(_image, sum, sqsum);
int sumStep = (int)(sum.step/sum.elemSize());
int sqsumStep = (int)(sqsum.step/sqsum.elemSize());
int tofs = hasTiltedFeatures ? sumStep*rn : 0;
CV_SUM_PTRS( p[0], p[1], p[2], p[3], sdata, normrect, sumStep );
CV_SUM_PTRS( pq[0], pq[1], pq[2], pq[3], sqdata, normrect, sqsumStep );
CV_SUM_OFS( nofs[0], nofs[1], nofs[2], nofs[3], 0, normrect, sumStep );
CV_SUM_OFS( nqofs[0], nqofs[1], nqofs[2], nqofs[3], 0, normrect, sqsumStep );
size_t fi, nfeatures = features.size();
size_t fi, nfeatures = features->size();
optfeatures->resize(nfeatures);
optfeaturesPtr = &(*optfeatures)[0];
const std::vector<Feature>& ff = *features;
for( fi = 0; fi < nfeatures; fi++ )
optfeaturesPtr[fi].updatePtrs( !featuresPtr[fi].tilted ? sum : tilted );
optfeaturesPtr[fi].setOffsets( ff[fi], sumStep, tofs );
return true;
}
@@ -555,10 +562,10 @@ bool HaarEvaluator::setWindow( Point pt )
pt.y + origWinSize.height >= sum.rows )
return false;
size_t pOffset = pt.y * (sum.step/sizeof(int)) + pt.x;
size_t pqOffset = pt.y * (sqsum.step/sizeof(double)) + pt.x;
int valsum = CALC_SUM(p, pOffset);
double valsqsum = CALC_SUM(pq, pqOffset);
const int* p = &sum.at<int>(pt);
const double* pq = &sqsum.at<double>(pt);
int valsum = CALC_SUM_OFS(nofs, p);
double valsqsum = CALC_SUM_OFS(nqofs, pq);
double nf = (double)normrect.area() * valsqsum - (double)valsum * valsum;
if( nf > 0. )
@@ -566,7 +573,7 @@ bool HaarEvaluator::setWindow( Point pt )
else
nf = 1.;
varianceNormFactor = 1./nf;
offset = (int)pOffset;
pwin = p;
return true;
}
@@ -613,8 +620,9 @@ Ptr<FeatureEvaluator> LBPEvaluator::clone() const
return ret;
}
bool LBPEvaluator::setImage( const Mat& image, Size _origWinSize )
bool LBPEvaluator::setImage( InputArray _image, Size _origWinSize )
{
Mat image = _image.getMat();
int rn = image.rows+1, cn = image.cols+1;
origWinSize = _origWinSize;
@@ -694,8 +702,9 @@ Ptr<FeatureEvaluator> HOGEvaluator::clone() const
return ret;
}
bool HOGEvaluator::setImage( const Mat& image, Size winSize )
bool HOGEvaluator::setImage( InputArray _image, Size winSize )
{
Mat image = _image.getMat();
int rows = image.rows + 1;
int cols = image.cols + 1;
origWinSize = winSize;
@@ -1011,11 +1020,11 @@ struct getRect { Rect operator ()(const CvAvgComp& e) const { return e.rect; } }
struct getNeighbors { int operator ()(const CvAvgComp& e) const { return e.neighbors; } };
bool CascadeClassifierImpl::detectSingleScale( const Mat& image, int stripCount, Size processingRectSize,
int stripSize, int yStep, double factor, std::vector<Rect>& candidates,
bool CascadeClassifierImpl::detectSingleScale( InputArray _image, Size processingRectSize,
int yStep, double factor, std::vector<Rect>& candidates,
std::vector<int>& levels, std::vector<double>& weights, bool outputRejectLevels )
{
if( !featureEvaluator->setImage( image, data.origWinSize ) )
if( !featureEvaluator->setImage( _image, data.origWinSize ) )
return false;
#if defined (LOG_CASCADE_STATISTIC)
@@ -1024,13 +1033,21 @@ bool CascadeClassifierImpl::detectSingleScale( const Mat& image, int stripCount,
Mat currentMask;
if (maskGenerator) {
Mat image = _image.getMat();
currentMask=maskGenerator->generateMask(image);
}
std::vector<Rect> candidatesVector;
std::vector<int> rejectLevels;
std::vector<double> levelWeights;
Mutex mtx;
int stripCount, stripSize;
const int PTS_PER_THREAD = 1000;
stripCount = ((processingRectSize.width/yStep)*(processingRectSize.height + yStep-1)/yStep + PTS_PER_THREAD/2)/PTS_PER_THREAD;
stripCount = std::min(std::max(stripCount, 1), 100);
stripSize = (((processingRectSize.height + stripCount - 1)/stripCount + yStep-1)/yStep)*yStep;
if( outputRejectLevels )
{
parallel_for_(Range(0, stripCount), CascadeClassifierInvoker( *this, processingRectSize, stripSize, yStep, factor,
@@ -1052,6 +1069,70 @@ bool CascadeClassifierImpl::detectSingleScale( const Mat& image, int stripCount,
return true;
}
bool CascadeClassifierImpl::ocl_detectSingleScale( InputArray _image, Size processingRectSize,
int yStep, double factor, std::vector<Rect>& candidates,
std::vector<int>&, std::vector<double>&, bool )
{
Ptr<HaarEvaluator> haar = featureEvaluator.dynamicCast<HaarEvaluator>();
if( haar.empty() )
return false;
if( cascadeKernel.empty() )
{
//cascadeKernel.create(")
if( cascadeKernel.empty() )
return false;
}
if( ustages.empty() )
{
#define UPLOAD_CASCADE_PART(NAME) \
Mat(1, (int)(data.NAME.size()*sizeof(data.NAME[0])), CV_8U, &data.NAME[0]).copyTo(u##NAME)
UPLOAD_CASCADE_PART(stages);
UPLOAD_CASCADE_PART(classifiers);
UPLOAD_CASCADE_PART(nodes);
UPLOAD_CASCADE_PART(leaves);
ufacepos.create();
}
haar->setUMat(_image, data.origWinSize, ugrayImage.size());
std::vector<UMat> bufs;
haar->getUMats(bufs);
CV_Assert(bufs.size() == 5);
size_t globalsize[] = { processingRectSize.width, processingRectSize.height };
if(!cascadeKernel.args(ocl::KernelArg::PtrReadOnly(bufs[0]), // sum
ocl::KernelArg::PtrReadOnly(bufs[1]), // sqsum
ocl::KernelArg::PtrReadOnly(bufs[2]), // optfeatures
// cascade classifier
ocl::KernelArg::PtrReadOnly(ustages),
ocl::KernelArg::PtrReadOnly(uclassifiers),
ocl::KernelArg::PtrReadOnly(unodes),
ocl::KernelArg::PtrReadOnly(uleaves),
ocl::KernelArg::WriteOnly(ufacepos), // positions
ocl::KernelArg::ReadWrite(umisc),
processingRectSize.width,
processingRectSize.height).run(2, globalsize, 0, false))
return false;
Mat facepos = ufacepos.getMat(ACCESS_READ);
const int* fptr = facepos.ptr<int>();
int nfaces = fptr[0];
for( i = 0; i < nfaces; i++ )
{
int pos = fptr[i+1];
int x =
candidates.push_back(Rect()
return false;
}
bool CascadeClassifierImpl::isOldFormatCascade() const
{
return !oldCascade.empty();
@@ -1097,28 +1178,56 @@ static void detectMultiScaleOldFormat( const Mat& image, Ptr<CvHaarClassifierCas
std::transform(vecAvgComp.begin(), vecAvgComp.end(), objects.begin(), getRect());
}
void CascadeClassifierImpl::detectMultiScaleNoGrouping( const Mat& image, std::vector<Rect>& candidates,
void CascadeClassifierImpl::detectMultiScaleNoGrouping( InputArray _image, std::vector<Rect>& candidates,
std::vector<int>& rejectLevels, std::vector<double>& levelWeights,
double scaleFactor, Size minObjectSize, Size maxObjectSize,
bool outputRejectLevels )
{
Size imgsz = _image.size();
int imgtype = _image.type();
Mat grayImage, imageBuffer;
candidates.clear();
if (maskGenerator)
maskGenerator->initializeMask(image);
rejectLevels.clear();
levelWeights.clear();
if( maxObjectSize.height == 0 || maxObjectSize.width == 0 )
maxObjectSize = image.size();
Mat grayImage = image;
if( grayImage.channels() > 1 )
maxObjectSize = imgsz;
bool use_ocl = ocl::useOpenCL() &&
getFeatureType() == FeatureEvaluator::HAAR &&
!isOldFormatCascade() &&
maskGenerator.empty() &&
!outputRejectLevels &&
tryOpenCL;
if( !use_ocl )
{
Mat temp;
cvtColor(grayImage, temp, COLOR_BGR2GRAY);
grayImage = temp;
}
Mat image = _image.getMat();
if (maskGenerator)
maskGenerator->initializeMask(image);
Mat imageBuffer(image.rows + 1, image.cols + 1, CV_8U);
grayImage = image;
if( CV_MAT_CN(imgtype) > 1 )
{
Mat temp;
cvtColor(grayImage, temp, COLOR_BGR2GRAY);
grayImage = temp;
}
imageBuffer.create(imgsz.height + 1, imgsz.width + 1, CV_8U);
}
else
{
UMat uimage = _image.getUMat();
if( CV_MAT_CN(imgtype) > 1 )
cvtColor(uimage, ugrayImage, COLOR_BGR2GRAY);
else
uimage.copyTo(ugrayImage);
uimageBuffer.create(imgsz.height + 1, imgsz.width + 1, CV_8U);
}
for( double factor = 1; ; factor *= scaleFactor )
{
@@ -1126,7 +1235,8 @@ void CascadeClassifierImpl::detectMultiScaleNoGrouping( const Mat& image, std::v
Size windowSize( cvRound(originalWindowSize.width*factor), cvRound(originalWindowSize.height*factor) );
Size scaledImageSize( cvRound( grayImage.cols/factor ), cvRound( grayImage.rows/factor ) );
Size processingRectSize( scaledImageSize.width - originalWindowSize.width, scaledImageSize.height - originalWindowSize.height );
Size processingRectSize( scaledImageSize.width - originalWindowSize.width,
scaledImageSize.height - originalWindowSize.height );
if( processingRectSize.width <= 0 || processingRectSize.height <= 0 )
break;
@@ -1134,10 +1244,7 @@ void CascadeClassifierImpl::detectMultiScaleNoGrouping( const Mat& image, std::v
break;
if( windowSize.width < minObjectSize.width || windowSize.height < minObjectSize.height )
continue;
Mat scaledImage( scaledImageSize, CV_8U, imageBuffer.data );
resize( grayImage, scaledImage, scaledImageSize, 0, 0, INTER_LINEAR );
int yStep;
if( getFeatureType() == cv::FeatureEvaluator::HOG )
{
@@ -1148,16 +1255,36 @@ void CascadeClassifierImpl::detectMultiScaleNoGrouping( const Mat& image, std::v
yStep = factor > 2. ? 1 : 2;
}
int stripCount, stripSize;
const int PTS_PER_THREAD = 1000;
stripCount = ((processingRectSize.width/yStep)*(processingRectSize.height + yStep-1)/yStep + PTS_PER_THREAD/2)/PTS_PER_THREAD;
stripCount = std::min(std::max(stripCount, 1), 100);
stripSize = (((processingRectSize.height + stripCount - 1)/stripCount + yStep-1)/yStep)*yStep;
if( !detectSingleScale( scaledImage, stripCount, processingRectSize, stripSize, yStep, factor, candidates,
rejectLevels, levelWeights, outputRejectLevels ) )
break;
if( use_ocl )
{
UMat uscaledImage(uimageBuffer, Rect(0, 0, scaledImageSize.width, scaledImageSize.height));
resize( ugrayImage, uscaledImage, scaledImageSize, 0, 0, INTER_LINEAR );
if( ocl_detectSingleScale( uscaledImage, processingRectSize, yStep, factor, candidates,
rejectLevels, levelWeights, outputRejectLevels ) )
continue;
/////// if the OpenCL branch has been executed but failed, fall back to CPU: /////
tryOpenCL = false; // for this cascade do not try OpenCL anymore
// since we may already have some partial results from OpenCL code (unlikely, but still),
// we just recursively call the function again, but with tryOpenCL==false it will
// go with CPU route, so there is no infinite recursion
detectMultiScaleNoGrouping( _image, candidates, rejectLevels, levelWeights,
scaleFactor, minObjectSize, maxObjectSize,
outputRejectLevels);
return;
}
else
{
Mat scaledImage( scaledImageSize, CV_8U, imageBuffer.data );
resize( grayImage, scaledImage, scaledImageSize, 0, 0, INTER_LINEAR );
if( !detectSingleScale( scaledImage, processingRectSize, yStep, factor, candidates,
rejectLevels, levelWeights, outputRejectLevels ) )
break;
}
}
}
@@ -1168,21 +1295,21 @@ void CascadeClassifierImpl::detectMultiScale( InputArray _image, std::vector<Rec
int flags, Size minObjectSize, Size maxObjectSize,
bool outputRejectLevels )
{
Mat image = _image.getMat();
CV_Assert( scaleFactor > 1 && image.depth() == CV_8U );
CV_Assert( scaleFactor > 1 && _image.depth() == CV_8U );
if( empty() )
return;
if( isOldFormatCascade() )
{
Mat image = _image.getMat();
std::vector<CvAvgComp> fakeVecAvgComp;
detectMultiScaleOldFormat( image, oldCascade, objects, rejectLevels, levelWeights, fakeVecAvgComp, scaleFactor,
minNeighbors, flags, minObjectSize, maxObjectSize, outputRejectLevels );
}
else
{
detectMultiScaleNoGrouping( image, objects, rejectLevels, levelWeights, scaleFactor, minObjectSize, maxObjectSize,
detectMultiScaleNoGrouping( _image, objects, rejectLevels, levelWeights, scaleFactor, minObjectSize, maxObjectSize,
outputRejectLevels );
const double GROUP_EPS = 0.2;
if( outputRejectLevels )
@@ -1346,8 +1473,15 @@ bool CascadeClassifierImpl::Data::read(const FileNode &root)
return true;
}
bool CascadeClassifierImpl::read_(const FileNode& root)
{
tryOpenCL = true;
cascadeKernel = ocl::Kernel();
ustages.release();
uclassifiers.release();
unodes.release();
uleaves.release();
if( !data.read(root) )
return false;