Maybe there is someone here that can help me fixing this issue: I am using Windows 10 64-bit, Visual Studio 2022, CMake 2.23.1, Anaconda with Python 3.9.7, and opencv-4.5.5 and opencv_contrib-4.5.5.
It worked that CUDA-module is available and computation works with CUDA support with my NVIDIA GPU - so far, everything is fine. But once I was trying to use the CascadeClassifier provided by OpenCV, it says:
I have checked the âopencvâ-folder in âsite-packagesâ as well, there is no data folder inside. I have tried it now several times, and it seems that the âdataâ-folder in âopencv-x.x.xâ-folder wonât get copied when configuring and building OpenCV with CUDA support.
Any advice to resolve this issue?
(I can post the build logs if you need it. CUDA works, I just donât know how to implement to data folder with OpenCV CUDA)
Yeah, it does not copy the âdataâ folder, where haarcascades_GPU is included as well
itâs even worse, the haarcascades_GPU data does not exist on current 3.x or 4.x branches. you must explicitly checkout the 2.4 branch to see or copy them
not really.
however, what i would like to see here, would be updated docs for the cuda classifier, and maybe a small tutorial / script, how to dl those cascades
itâs even worse, the haarcascades_GPU data does not exist on current 3.x or 4.x branches. you must explicitly checkout the 2.4 branch to see or copy them
Phew⌠okay, I see - thanks.
however, what i would like to see here, would be updated docs for the cuda classifier, and maybe a small tutorial / script, how to dl those cascades
Yeah, I absolutely agree. Apart from that, even the CUDA support for OpenCV is quite difficult to be set up from my pov as a user. I think it would be way easier if CUDA support would be auto-integrated into OpenCV and detected if there is a supported GPU detected (e.g., like it works in TensorFlow 2.x).