Commit a64d3c17 authored by Andrey Pavlenko's avatar Andrey Pavlenko Committed by OpenCV Buildbot

Merge pull request #2155 from vbystricky:ocl_calcOpticalFlowPyrLK

parents 1db8aa86 bb09d44e
......@@ -60,6 +60,7 @@ class CV_EXPORTS Program;
class CV_EXPORTS ProgramSource2;
class CV_EXPORTS Queue;
class CV_EXPORTS PlatformInfo2;
class CV_EXPORTS Image2D;
class CV_EXPORTS Device
{
......@@ -89,6 +90,7 @@ public:
String vendor() const;
String OpenCL_C_Version() const;
String OpenCLVersion() const;
String deviceVersion() const;
String driverVersion() const;
void* ptr() const;
......@@ -325,6 +327,7 @@ public:
const String& buildopts, String* errmsg=0);
int set(int i, const void* value, size_t sz);
int set(int i, const Image2D& image2D);
int set(int i, const UMat& m);
int set(int i, const KernelArg& arg);
template<typename _Tp> int set(int i, const _Tp& value)
......@@ -574,6 +577,19 @@ CV_EXPORTS const char* typeToStr(int t);
CV_EXPORTS const char* memopTypeToStr(int t);
CV_EXPORTS void getPlatfomsInfo(std::vector<PlatformInfo2>& platform_info);
class CV_EXPORTS Image2D
{
public:
Image2D();
Image2D(const UMat &src);
~Image2D();
void* ptr() const;
protected:
struct Impl;
Impl* p;
};
}}
#endif
......@@ -821,6 +821,7 @@ OCL_FUNC_P(cl_mem, clCreateSubBuffer,
const void * buffer_create_info,
cl_int * errcode_ret),
(buffer, flags, buffer_create_type, buffer_create_info, errcode_ret))
*/
OCL_FUNC_P(cl_mem, clCreateImage,
(cl_context context,
......@@ -831,6 +832,18 @@ OCL_FUNC_P(cl_mem, clCreateImage,
cl_int * errcode_ret),
(context, flags, image_format, image_desc, host_ptr, errcode_ret))
OCL_FUNC_P(cl_mem, clCreateImage2D,
(cl_context context,
cl_mem_flags flags,
const cl_image_format * image_format,
size_t image_width,
size_t image_height,
size_t image_row_pitch,
void * host_ptr,
cl_int *errcode_ret),
(context, flags, image_format, image_width, image_height, image_row_pitch, host_ptr, errcode_ret))
/*
OCL_FUNC(cl_int, clGetSupportedImageFormats,
(cl_context context,
cl_mem_flags flags,
......@@ -945,21 +958,26 @@ OCL_FUNC(cl_int, clEnqueueCopyImageToBuffer,
cl_event * event),
(command_queue, src_image, dst_buffer, src_origin, region, dst_offset,
num_events_in_wait_list, event_wait_list, event))
*/
OCL_FUNC(cl_int, clEnqueueCopyBufferToImage,
(cl_command_queue command_queue,
cl_mem src_buffer,
cl_mem dst_image,
size_t src_offset,
const size_t * dst_origin[3],
const size_t * region[3],
const size_t dst_origin[3],
const size_t region[3],
cl_uint num_events_in_wait_list,
const cl_event * event_wait_list,
cl_event * event),
(command_queue, src_buffer, dst_image, src_offset, dst_origin,
region, num_events_in_wait_list, event_wait_list, event))
OCL_FUNC(cl_int, clFlush,
(cl_command_queue command_queue),
(command_queue))
/*
OCL_FUNC_P(void*, clEnqueueMapImage,
(cl_command_queue command_queue,
cl_mem image,
......@@ -976,7 +994,9 @@ OCL_FUNC_P(void*, clEnqueueMapImage,
(command_queue, image, blocking_map, map_flags, origin, region,
image_row_pitch, image_slice_pitch, num_events_in_wait_list,
event_wait_list, event, errcode_ret))
*/
/*
OCL_FUNC(cl_int, clRetainProgram, (cl_program program), (program))
OCL_FUNC(cl_int, clGetKernelInfo,
......@@ -1705,6 +1725,9 @@ String Device::OpenCL_C_Version() const
String Device::OpenCLVersion() const
{ return p ? p->getStrProp(CL_DEVICE_EXTENSIONS) : String(); }
String Device::deviceVersion() const
{ return p ? p->getStrProp(CL_DEVICE_VERSION) : String(); }
String Device::driverVersion() const
{ return p ? p->getStrProp(CL_DRIVER_VERSION) : String(); }
......@@ -2689,6 +2712,12 @@ int Kernel::set(int i, const void* value, size_t sz)
return i+1;
}
int Kernel::set(int i, const Image2D& image2D)
{
cl_mem h = (cl_mem)image2D.ptr();
return set(i, &h, sizeof(h));
}
int Kernel::set(int i, const UMat& m)
{
return set(i, KernelArg(KernelArg::READ_WRITE, (UMat*)&m, 0, 0));
......@@ -3785,4 +3814,151 @@ const char* convertTypeStr(int sdepth, int ddepth, int cn, char* buf)
return buf;
}
///////////////////////////////////////////////////////////////////////////////////////////////
// deviceVersion has format
// OpenCL<space><major_version.minor_version><space><vendor-specific information>
// by specification
// http://www.khronos.org/registry/cl/sdk/1.1/docs/man/xhtml/clGetDeviceInfo.html
// http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clGetDeviceInfo.html
static void parseDeviceVersion(const String &deviceVersion, int &major, int &minor)
{
major = minor = 0;
if (10 >= deviceVersion.length())
return;
const char *pstr = deviceVersion.c_str();
if (0 != strncmp(pstr, "OpenCL ", 7))
return;
size_t ppos = deviceVersion.find('.', 7);
if (String::npos == ppos)
return;
String temp = deviceVersion.substr(7, ppos - 7);
major = atoi(temp.c_str());
temp = deviceVersion.substr(ppos + 1);
minor = atoi(temp.c_str());
}
struct Image2D::Impl
{
Impl(const UMat &src)
{
handle = 0;
refcount = 1;
init(src);
}
~Impl()
{
if (handle)
clReleaseMemObject(handle);
}
void init(const UMat &src)
{
cl_image_format format;
int err;
int depth = src.depth();
int channels = src.channels();
switch(depth)
{
case CV_8U:
format.image_channel_data_type = CL_UNSIGNED_INT8;
break;
case CV_32S:
format.image_channel_data_type = CL_UNSIGNED_INT32;
break;
case CV_32F:
format.image_channel_data_type = CL_FLOAT;
break;
default:
CV_Error(-1, "Image forma is not supported");
break;
}
switch(channels)
{
case 1:
format.image_channel_order = CL_R;
break;
case 3:
format.image_channel_order = CL_RGB;
break;
case 4:
format.image_channel_order = CL_RGBA;
break;
default:
CV_Error(-1, "Image format is not supported");
break;
}
#ifdef CL_VERSION_1_2
//this enables backwards portability to
//run on OpenCL 1.1 platform if library binaries are compiled with OpenCL 1.2 support
int minor, major;
parseDeviceVersion(Device::getDefault().deviceVersion(), major, minor);
if ((1 < major) || ((1 == major) && (2 <= minor)))
{
cl_image_desc desc;
desc.image_type = CL_MEM_OBJECT_IMAGE2D;
desc.image_width = src.cols;
desc.image_height = src.rows;
desc.image_depth = 0;
desc.image_array_size = 1;
desc.image_row_pitch = 0;
desc.image_slice_pitch = 0;
desc.buffer = NULL;
desc.num_mip_levels = 0;
desc.num_samples = 0;
handle = clCreateImage((cl_context)Context2::getDefault().ptr(), CL_MEM_READ_WRITE, &format, &desc, NULL, &err);
}
else
#endif
{
handle = clCreateImage2D((cl_context)Context2::getDefault().ptr(), CL_MEM_READ_WRITE, &format, src.cols, src.rows, 0, NULL, &err);
}
size_t origin[] = { 0, 0, 0 };
size_t region[] = { src.cols, src.rows, 1 };
cl_mem devData;
if (!src.isContinuous())
{
devData = clCreateBuffer((cl_context)Context2::getDefault().ptr(), CL_MEM_READ_ONLY, src.cols * src.rows * src.elemSize(), NULL, NULL);
const size_t roi[3] = {src.cols * src.elemSize(), src.rows, 1};
clEnqueueCopyBufferRect((cl_command_queue)Queue::getDefault().ptr(), (cl_mem)src.handle(ACCESS_READ), devData, origin, origin,
roi, src.step, 0, src.cols * src.elemSize(), 0, 0, NULL, NULL);
clFlush((cl_command_queue)Queue::getDefault().ptr());
}
else
{
devData = (cl_mem)src.handle(ACCESS_READ);
}
clEnqueueCopyBufferToImage((cl_command_queue)Queue::getDefault().ptr(), devData, handle, 0, origin, region, 0, NULL, 0);
if (!src.isContinuous())
{
clFlush((cl_command_queue)Queue::getDefault().ptr());
clReleaseMemObject(devData);
}
}
IMPLEMENT_REFCOUNTABLE();
cl_mem handle;
};
Image2D::Image2D()
{
p = NULL;
}
Image2D::Image2D(const UMat &src)
{
p = new Impl(src);
}
Image2D::~Image2D()
{
if (p)
p->release();
}
void* Image2D::ptr() const
{
return p ? p->handle : 0;
}
}}
/*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"
using std::tr1::make_tuple;
#ifdef HAVE_OPENCL
namespace cvtest {
namespace ocl {
///////////// FarnebackOpticalFlow ////////////////////////
CV_ENUM(farneFlagType, 0, OPTFLOW_FARNEBACK_GAUSSIAN)
typedef tuple< int > PyrLKOpticalFlowParams;
typedef TestBaseWithParam<PyrLKOpticalFlowParams> PyrLKOpticalFlowFixture;
OCL_PERF_TEST_P(PyrLKOpticalFlowFixture, PyrLKOpticalFlow,
::testing::Values(1000, 2000, 4000)
)
{
Mat frame0 = imread(getDataPath("gpu/opticalflow/rubberwhale1.png"), cv::IMREAD_GRAYSCALE);
ASSERT_FALSE(frame0.empty()) << "can't load rubberwhale1.png";
Mat frame1 = imread(getDataPath("gpu/opticalflow/rubberwhale2.png"), cv::IMREAD_GRAYSCALE);
ASSERT_FALSE(frame1.empty()) << "can't load rubberwhale2.png";
UMat uFrame0; frame0.copyTo(uFrame0);
UMat uFrame1; frame1.copyTo(uFrame1);
const Size winSize = Size(21, 21);
const int maxLevel = 3;
const TermCriteria criteria = TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, 0.01);
const int flags = 0;
const float minEigThreshold = 1e-4f;
const double eps = 1.0;
const PyrLKOpticalFlowParams params = GetParam();
const int pointsCount = get<0>(params);
vector<Point2f> pts, nextPts;
vector<unsigned char> status;
vector<float> err;
goodFeaturesToTrack(frame0, pts, pointsCount, 0.01, 0.0);
Mat ptsMat(1, static_cast<int>(pts.size()), CV_32FC2, (void *)&pts[0]);
declare.in(uFrame0, uFrame1, WARMUP_READ);
UMat uNextPts, uStatus, uErr;
OCL_TEST_CYCLE()
cv::calcOpticalFlowPyrLK(uFrame0, uFrame1, pts, uNextPts, uStatus, uErr, winSize, maxLevel, criteria, flags, minEigThreshold);
SANITY_CHECK(uNextPts, eps);
}
} } // namespace cvtest::ocl
#endif // HAVE_OPENCL
\ No newline at end of file
This diff is collapsed.
This diff is collapsed.
/*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, 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.
//
// 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 cvtest {
namespace ocl {
/////////////////////////////////////////////////////////////////////////////////////////////////
// PyrLKOpticalFlow
PARAM_TEST_CASE(PyrLKOpticalFlow, int, int)
{
Size winSize;
int maxLevel;
TermCriteria criteria;
int flags;
double minEigThreshold;
virtual void SetUp()
{
winSize = Size(GET_PARAM(0), GET_PARAM(0));
maxLevel = GET_PARAM(1);
criteria = TermCriteria(TermCriteria::COUNT+TermCriteria::EPS, 30, 0.01);
flags = 0;
minEigThreshold = 1e-4f;
}
};
OCL_TEST_P(PyrLKOpticalFlow, Mat)
{
static const int npoints = 1000;
static const float eps = 0.03f;
cv::Mat frame0 = readImage("optflow/RubberWhale1.png", cv::IMREAD_GRAYSCALE);
ASSERT_FALSE(frame0.empty());
UMat umatFrame0; frame0.copyTo(umatFrame0);
cv::Mat frame1 = readImage("optflow/RubberWhale2.png", cv::IMREAD_GRAYSCALE);
ASSERT_FALSE(frame1.empty());
UMat umatFrame1; frame1.copyTo(umatFrame1);
std::vector<cv::Point2f> pts;
cv::goodFeaturesToTrack(frame0, pts, npoints, 0.01, 0.0);
std::vector<cv::Point2f> cpuNextPts;
std::vector<unsigned char> cpuStatusCPU;
std::vector<float> cpuErr;
OCL_OFF(cv::calcOpticalFlowPyrLK(frame0, frame1, pts, cpuNextPts, cpuStatusCPU, cpuErr, winSize, maxLevel, criteria, flags, minEigThreshold));
UMat umatNextPts, umatStatus, umatErr;
OCL_ON(cv::calcOpticalFlowPyrLK(umatFrame0, umatFrame1, pts, umatNextPts, umatStatus, umatErr, winSize, maxLevel, criteria, flags, minEigThreshold));
std::vector<cv::Point2f> nextPts; umatNextPts.reshape(2, 1).copyTo(nextPts);
std::vector<unsigned char> status; umatStatus.reshape(1, 1).copyTo(status);
std::vector<float> err; umatErr.reshape(1, 1).copyTo(err);
ASSERT_EQ(cpuNextPts.size(), nextPts.size());
ASSERT_EQ(cpuStatusCPU.size(), status.size());
size_t mistmatch = 0;
for (size_t i = 0; i < nextPts.size(); ++i)
{
if (status[i] != cpuStatusCPU[i])
{
++mistmatch;
continue;
}
if (status[i])
{
cv::Point2i a = nextPts[i];
cv::Point2i b = cpuNextPts[i];
bool eq = std::abs(a.x - b.x) < 1 && std::abs(a.y - b.y) < 1;
float errdiff = 0.0f;
if (!eq || errdiff > 1e-1)
++mistmatch;
}
}
double bad_ratio = static_cast<double>(mistmatch) / (nextPts.size());
ASSERT_LE(bad_ratio, eps);
}
OCL_INSTANTIATE_TEST_CASE_P(Video, PyrLKOpticalFlow,
Combine(
Values(21, 25),
Values(3, 5)
)
);
} } // namespace cvtest::ocl
#endif // HAVE_OPENCL
\ No newline at end of file
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment