Skip to content

opencv-python-cuda 5.0.0-dev1

Pre-release
Pre-release

Choose a tag to compare

@Breakthrough Breakthrough released this 07 Nov 02:07

Preview build of OpenCV 5.0.0 with CUDA support, including hardware video encode/decode (NVDEC/NVENC).

OpenCV 5 is a major release with breaking API changes. Review the upstream OpenCV change logs before upgrading from 4.x.

Installation

From the package index:

pip install "opencv-python-cuda==5.0.0+945546e" --extra-index-url https://breakthrough.github.io/opencv-python-cuda/packages/

Or download the wheel below and install it directly:

pip install opencv_python_cuda-5.0.0+945546e-cp37-abi3-win_amd64.whl

Make sure to uninstall any existing OpenCV packages first. This package can be uninstalled via pip uninstall opencv-python-cuda. To stay on OpenCV 4.x when installing from the package index, use "opencv-python-cuda<5".

Build/Version Information

  • Build Run
  • OpenCV 5.0.0 with all features excluding non-free
  • Nvidia CUDA Toolkit v12.9.2
  • Nvidia CuDNN 9.28.0.42
  • Nvidia Video Codec SDK 13.0.37 (hardware video decode/encode)
  • Device compatibility: Maxwell through Blackwell (SASS: 5.0, 5.2, 6.0, 6.1, 7.0, 7.5, 8.0, 8.6, 8.9, 9.0, 10.0; PTX: 10.0)

System Requirements

  • Windows 7 64-bit or newer
  • Python 3.7 or newer
  • Nvidia GPU that is Maxwell-class or newer (GeForce 900-series)
  • Nvidia display driver must be installed: cv2 links against nvcuvid.dll / nvEncodeAPI64.dll, which ship with the driver, so import cv2 fails on machines without it

Changes from 5.0.0-dev0

  • Fixed the DNN CUDA backend. Using cv2.dnn with DNN_BACKEND_CUDA (including cv2.dnn.getAvailableTargets(cv2.dnn.DNN_BACKEND_CUDA)) terminated the Python process with Cannot load symbol cudnnGetVersion, because cuDNN 9 could not find its own sub-libraries in the cv2 package directory. Earlier releases that bundle cuDNN 9 (including 5.0.0-dev0) are affected; as a workaround for those, add the installed cv2 package directory to PATH before importing cv2.
  • Updated Nvidia CUDA Toolkit from 12.9.1 to 12.9.2
  • Updated Nvidia CuDNN from 9.18.1.3 to 9.28.0.42
  • CI now checks that cuDNN's sub-libraries load from the package (tools/check_cudnn_loading.py)