GPU Computer Vision Library

Multi-platform (Windows, Unix and Mac OS) C++/OpenCL/CUDA library with several Computer Vision algorithms.

Building:

cmake -G "Unix Makefiles" -DBUILD_OPENCL=ON -DBUILD_CUDA=ON -DBUILD_TESTS=ON .
make

Example:

#include <gcvl/blockmatching.h>
#include <gcvl/opencl/oclcore.h>
#include <gcvl/opencl/oclblockmatching.h>
#include <gcvl/cuda/cudacore.h>
#include <gcvl/cuda/cudablockmatching.h>

//argv[1] -> path to left image   argv[2] -> path to right image
int main(int argc, char *argv[]) {

  int dim = 5, maxDisp = 16;
  bool norm = true;
  std::unique_ptr<unsigned char[]> output;
  
  gcvl::BlockMatching bmCPU(argv[1], argv[2], output);
  bmCPU.setAggDim(dim);
  bmCPU.setMaxDisp(maxDisp);
  bmCPU.setNormalize(norm);
  bmCPU.compute();
  
  gcvl::opencl::Core coreOCL;
  gcvl::opencl::BlockMatching bmOCL(coreOCL, argv[1], argv[2], output);
  bmOCL.setAggDim(dim);
  bmOCL.setMaxDisp(maxDisp);
  bmOCL.setNormalize(norm);
  bmOCL.compute();
  
  gcvl::cuda::Core coreCUDA;
  gcvl::cuda::BlockMatching bmCUDA(coreCUDA, argv[1], argv[2], output);
  bmCUDA.setAggDim(dim);
  bmCUDA.setMaxDisp(maxDisp);
  bmCUDA.setNormalize(norm);
  bmCUDA.compute();
  
  return 0;
  
}

Documentation:

Doxygen documentation is available for GCVL here.