Problem building OpenCV with CUDA support from source files with Cmake (Python)

[cudawarped]

@Eljais

However, after trying numerous times, I always encounter problems. I believe it has to do with the specific Cuda/CUDNN version, my compute capability (8.6), the used version of Visual Studio and the matching OpenCV and OpenCV Contrib version.

What error/warning/output makes you think its a CUDA/cuDNN issue?

@Eljais

Using the latest versions of OpenCV (4.8.0) and Cuda/CUDNN (12.3 or 12.2) did not work, as after building and generating the source files and running the cmake command in Command Prompt (“C:\Program Files\CMake\bin\cmake.exe” --build “C:\OpenCV_Build\build” --target INSTALL --config Release) there still did not appear a CV2 folder in my anaconda3 sitepackages, neither did the file cv2.cp38-win_amd64.pyd appear in either the build folder or the anaconda folder.

This implies you haven’t specified the pythonpaths correctly. I would advise starting again and using the command line as described in this guide

Build OpenCV (including Python) with CUDA on Windows: Comprehensive Guide...

then inspecting the CMake output before building to make sure that python3 will be built and that you have something similar to

– Python 3:
– Interpreter: C:/Users/username/mambaforge/python.exe (ver 3.9.16)
– Libraries: C:/Users/username/mambaforge/libs/python39.lib (ver 3.9.16)
– numpy: C:/Users/username/mambaforge/Lib/site-packages/numpy/core/include (ver 1.23.5)
– install path: C:/Users/username/mambaforge/Lib/site-packages/cv2/python-3.9