I'm trying to connect up OpenCV and NVIDIA's own libraries in Python as fast as I can, and I've hit a bit of a snag. I'd like to make a GpuMat, using the constructor:
GpuMat(int rows, int cols, int type, void* data, size_t step = Mat::AUTO_STEP); in order to just assign the memory location instead of copying anything (there's some is-the-data-valid questions there, but assume that's resolved). But, the macro CVWRAP is conspicuously missing from that particular constructor, so it doesn't get built into the python library. Before I jump into the rabbit hole, does anyone know why? Can I just add CVWRAP and be on my way, or is there a gotcha waiting for me?
The relevant line is include/opencv2/core/cuda.hpp:137
And the definition: modules/core/src/cuda_gpu_mat.cpp:56
I'm on Windows, if it matters. Alternatively, is this functionality already there somewhere else, and I'm missing it?