Hi everyone, I am trying to use the HoG-GPU implemented class of OpenCV for detecting objects other than people, so I have trained my HoG descriptors using CPU first. But due to speed matters I needed to switch to detect using GPU, mostly getting help from here.
1- Is it so that the GPU class of HoG is implemented just for linear Kernel? I didn't manage to load RBF saved vectors!? 2- Unfortunately the results of the GPU vs CPU detection are totally different. I've read this also in other posts. Does anyone have experiences in this case? (
//gpu_hog settings:
Size win_size(32, 32); Size block_size(16, 16); Size block_stride(8, 8); Size cell_size(8, 8); int nbins = 18;
Thanks