Merge pull request #14888 from vchiluka5:NVIDIA_Optical_Flow

pull/15000/head
Alexander Alekhin 5 years ago committed by GitHub
commit 8408587341
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
  1. 24
      CMakeLists.txt
  2. 49
      cmake/OpenCVDetectCUDA.cmake
  3. 240
      samples/gpu/optical_flow.cpp

@ -918,30 +918,6 @@ configure_file("${OpenCV_SOURCE_DIR}/cmake/templates/custom_hal.hpp.in" "${CMAKE
unset(_hal_includes)
# ----------------------------------------------------------------------------
# Add CUDA libraries (needed for apps/tools, samples)
# ----------------------------------------------------------------------------
if(HAVE_CUDA)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_LIBRARIES} ${CUDA_npp_LIBRARY})
if(HAVE_CUBLAS)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_cublas_LIBRARY})
endif()
if(HAVE_CUDNN)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDNN_LIBRARIES})
endif()
if(HAVE_CUFFT)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_cufft_LIBRARY})
endif()
foreach(p ${CUDA_LIBS_PATH})
if(MSVC AND CMAKE_GENERATOR MATCHES "Ninja|JOM")
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CMAKE_LIBRARY_PATH_FLAG}"${p}")
else()
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CMAKE_LIBRARY_PATH_FLAG}${p})
endif()
endforeach()
endif()
# ----------------------------------------------------------------------------
# Code trace support
# ----------------------------------------------------------------------------

@ -337,3 +337,52 @@ if(HAVE_CUDA)
ocv_convert_to_lib_name(CUDA_cufft_LIBRARY ${CUDA_cufft_LIBRARY})
endif()
endif()
# ----------------------------------------------------------------------------
# Add CUDA libraries (needed for apps/tools, samples)
# ----------------------------------------------------------------------------
if(HAVE_CUDA)
# details: https://github.com/NVIDIA/nvidia-docker/issues/775
if(" ${CUDA_CUDA_LIBRARY}" MATCHES "/stubs/libcuda.so" AND NOT OPENCV_SKIP_CUDA_STUB_WORKAROUND)
set(CUDA_STUB_ENABLED_LINK_WORKAROUND 1)
if(EXISTS "${CUDA_CUDA_LIBRARY}" AND NOT OPENCV_SKIP_CUDA_STUB_WORKAROUND_RPATH_LINK)
set(CUDA_STUB_TARGET_PATH "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/")
execute_process(COMMAND ${CMAKE_COMMAND} -E create_symlink "${CUDA_CUDA_LIBRARY}" "${CUDA_STUB_TARGET_PATH}/libcuda.so.1"
RESULT_VARIABLE CUDA_STUB_SYMLINK_RESULT)
if(NOT CUDA_STUB_SYMLINK_RESULT EQUAL 0)
execute_process(COMMAND ${CMAKE_COMMAND} -E copy_if_different "${CUDA_CUDA_LIBRARY}" "${CUDA_STUB_TARGET_PATH}/libcuda.so.1"
RESULT_VARIABLE CUDA_STUB_COPY_RESULT)
if(NOT CUDA_STUB_COPY_RESULT EQUAL 0)
set(CUDA_STUB_ENABLED_LINK_WORKAROUND 0)
endif()
endif()
if(CUDA_STUB_ENABLED_LINK_WORKAROUND)
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -Wl,-rpath-link,\"${CUDA_STUB_TARGET_PATH}\"")
endif()
else()
set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -Wl,--allow-shlib-undefined")
endif()
if(NOT CUDA_STUB_ENABLED_LINK_WORKAROUND)
message(WARNING "CUDA: workaround for stubs/libcuda.so.1 is not applied")
endif()
endif()
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_LIBRARIES} ${CUDA_CUDA_LIBRARY} ${CUDA_npp_LIBRARY})
if(HAVE_CUBLAS)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_cublas_LIBRARY})
endif()
if(HAVE_CUDNN)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDNN_LIBRARIES})
endif()
if(HAVE_CUFFT)
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CUDA_cufft_LIBRARY})
endif()
foreach(p ${CUDA_LIBS_PATH})
if(MSVC AND CMAKE_GENERATOR MATCHES "Ninja|JOM")
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CMAKE_LIBRARY_PATH_FLAG}"${p}")
else()
set(OPENCV_LINKER_LIBS ${OPENCV_LINKER_LIBS} ${CMAKE_LIBRARY_PATH_FLAG}${p})
endif()
endforeach()
endif()

@ -1,240 +0,0 @@
#include <iostream>
#include <fstream>
#include "opencv2/core.hpp"
#include <opencv2/core/utility.hpp>
#include "opencv2/highgui.hpp"
#include "opencv2/cudaoptflow.hpp"
#include "opencv2/cudaarithm.hpp"
using namespace std;
using namespace cv;
using namespace cv::cuda;
inline bool isFlowCorrect(Point2f u)
{
return !cvIsNaN(u.x) && !cvIsNaN(u.y) && fabs(u.x) < 1e9 && fabs(u.y) < 1e9;
}
static Vec3b computeColor(float fx, float fy)
{
static bool first = true;
// relative lengths of color transitions:
// these are chosen based on perceptual similarity
// (e.g. one can distinguish more shades between red and yellow
// than between yellow and green)
const int RY = 15;
const int YG = 6;
const int GC = 4;
const int CB = 11;
const int BM = 13;
const int MR = 6;
const int NCOLS = RY + YG + GC + CB + BM + MR;
static Vec3i colorWheel[NCOLS];
if (first)
{
int k = 0;
for (int i = 0; i < RY; ++i, ++k)
colorWheel[k] = Vec3i(255, 255 * i / RY, 0);
for (int i = 0; i < YG; ++i, ++k)
colorWheel[k] = Vec3i(255 - 255 * i / YG, 255, 0);
for (int i = 0; i < GC; ++i, ++k)
colorWheel[k] = Vec3i(0, 255, 255 * i / GC);
for (int i = 0; i < CB; ++i, ++k)
colorWheel[k] = Vec3i(0, 255 - 255 * i / CB, 255);
for (int i = 0; i < BM; ++i, ++k)
colorWheel[k] = Vec3i(255 * i / BM, 0, 255);
for (int i = 0; i < MR; ++i, ++k)
colorWheel[k] = Vec3i(255, 0, 255 - 255 * i / MR);
first = false;
}
const float rad = sqrt(fx * fx + fy * fy);
const float a = atan2(-fy, -fx) / (float) CV_PI;
const float fk = (a + 1.0f) / 2.0f * (NCOLS - 1);
const int k0 = static_cast<int>(fk);
const int k1 = (k0 + 1) % NCOLS;
const float f = fk - k0;
Vec3b pix;
for (int b = 0; b < 3; b++)
{
const float col0 = colorWheel[k0][b] / 255.0f;
const float col1 = colorWheel[k1][b] / 255.0f;
float col = (1 - f) * col0 + f * col1;
if (rad <= 1)
col = 1 - rad * (1 - col); // increase saturation with radius
else
col *= .75; // out of range
pix[2 - b] = static_cast<uchar>(255.0 * col);
}
return pix;
}
static void drawOpticalFlow(const Mat_<float>& flowx, const Mat_<float>& flowy, Mat& dst, float maxmotion = -1)
{
dst.create(flowx.size(), CV_8UC3);
dst.setTo(Scalar::all(0));
// determine motion range:
float maxrad = maxmotion;
if (maxmotion <= 0)
{
maxrad = 1;
for (int y = 0; y < flowx.rows; ++y)
{
for (int x = 0; x < flowx.cols; ++x)
{
Point2f u(flowx(y, x), flowy(y, x));
if (!isFlowCorrect(u))
continue;
maxrad = max(maxrad, sqrt(u.x * u.x + u.y * u.y));
}
}
}
for (int y = 0; y < flowx.rows; ++y)
{
for (int x = 0; x < flowx.cols; ++x)
{
Point2f u(flowx(y, x), flowy(y, x));
if (isFlowCorrect(u))
dst.at<Vec3b>(y, x) = computeColor(u.x / maxrad, u.y / maxrad);
}
}
}
static void showFlow(const char* name, const GpuMat& d_flow)
{
GpuMat planes[2];
cuda::split(d_flow, planes);
Mat flowx(planes[0]);
Mat flowy(planes[1]);
Mat out;
drawOpticalFlow(flowx, flowy, out, 10);
imshow(name, out);
}
int main(int argc, const char* argv[])
{
string filename1, filename2;
if (argc < 3)
{
cerr << "Usage : " << argv[0] << " <frame0> <frame1>" << endl;
filename1 = "../data/basketball1.png";
filename2 = "../data/basketball2.png";
}
else
{
filename1 = argv[1];
filename2 = argv[2];
}
Mat frame0 = imread(filename1, IMREAD_GRAYSCALE);
Mat frame1 = imread(filename2, IMREAD_GRAYSCALE);
if (frame0.empty())
{
cerr << "Can't open image [" << filename1 << "]" << endl;
return -1;
}
if (frame1.empty())
{
cerr << "Can't open image [" << filename2 << "]" << endl;
return -1;
}
if (frame1.size() != frame0.size())
{
cerr << "Images should be of equal sizes" << endl;
return -1;
}
GpuMat d_frame0(frame0);
GpuMat d_frame1(frame1);
GpuMat d_flow(frame0.size(), CV_32FC2);
Ptr<cuda::BroxOpticalFlow> brox = cuda::BroxOpticalFlow::create(0.197f, 50.0f, 0.8f, 10, 77, 10);
Ptr<cuda::DensePyrLKOpticalFlow> lk = cuda::DensePyrLKOpticalFlow::create(Size(7, 7));
Ptr<cuda::FarnebackOpticalFlow> farn = cuda::FarnebackOpticalFlow::create();
Ptr<cuda::OpticalFlowDual_TVL1> tvl1 = cuda::OpticalFlowDual_TVL1::create();
{
GpuMat d_frame0f;
GpuMat d_frame1f;
d_frame0.convertTo(d_frame0f, CV_32F, 1.0 / 255.0);
d_frame1.convertTo(d_frame1f, CV_32F, 1.0 / 255.0);
const int64 start = getTickCount();
brox->calc(d_frame0f, d_frame1f, d_flow);
const double timeSec = (getTickCount() - start) / getTickFrequency();
cout << "Brox : " << timeSec << " sec" << endl;
showFlow("Brox", d_flow);
}
{
const int64 start = getTickCount();
lk->calc(d_frame0, d_frame1, d_flow);
const double timeSec = (getTickCount() - start) / getTickFrequency();
cout << "LK : " << timeSec << " sec" << endl;
showFlow("LK", d_flow);
}
{
const int64 start = getTickCount();
farn->calc(d_frame0, d_frame1, d_flow);
const double timeSec = (getTickCount() - start) / getTickFrequency();
cout << "Farn : " << timeSec << " sec" << endl;
showFlow("Farn", d_flow);
}
{
const int64 start = getTickCount();
tvl1->calc(d_frame0, d_frame1, d_flow);
const double timeSec = (getTickCount() - start) / getTickFrequency();
cout << "TVL1 : " << timeSec << " sec" << endl;
showFlow("TVL1", d_flow);
}
imshow("Frame 0", frame0);
imshow("Frame 1", frame1);
waitKey();
return 0;
}
Loading…
Cancel
Save