Revision history [back]

This is done at least since 2.4.7. If you build OpenCV with CUDA flags with driver installed, opencv_traincascade will use CUDA and you should see a significant speedup:

$ otool -L bin/opencv_traincascade
bin/opencv_traincascade:
    @rpath/libcudart.5.5.dylib (compatibility version 0.0.0, current version 5.5.28)
...
    @rpath/libcublas.5.5.dylib (compatibility version 0.0.0, current version 5.5.28)
    @rpath/libcufft.5.5.dylib (compatibility version 0.0.0, current version 5.5.28)
...

This is done at least since 2.4.7. If you build OpenCV with CUDA flags with driver installed, opencv_traincascade will use CUDA and you should see a significant speedup:CUDA:

$ otool -L bin/opencv_traincascade
bin/opencv_traincascade:
    @rpath/libcudart.5.5.dylib (compatibility version 0.0.0, current version 5.5.28)
...
    @rpath/libcublas.5.5.dylib (compatibility version 0.0.0, current version 5.5.28)
    @rpath/libcufft.5.5.dylib (compatibility version 0.0.0, current version 5.5.28)
...