Beta

CV-CUDA 0.3.0 is the next open-source release of the project.

Release Highlights

CV-CUDA 0.3.0 includes the following key features:

  • 6 new computer vision operators

    • Adaptive Thresholding, Bounding Box, Bounding Box Blur, Non-Maximum Suppression, Remap, and Thresholding.

  • Additional compiler support for GCC 9+ (unit tests build with GCC 11+)

  • Compatibility with CUDA Toolkit 11.2+ (unit tests build with Toolkit 11.7+)

  • New sample applications

    • Object Detection - This pipeline is based on Peoplenet Tao model which detects Persons, Bags and Face - Supports TensorRT Inference Backend

    • Segmentation with NVIDIA Triton Backend - Uses VPF decoder/encoder on client side and runs segmentation pipeline on Triton server

  • Improved testing

Compatibility

CV-CUDA has been tested on the following compute stack

  • Ubuntu x86_64: 18.04, 20.04, 22.04

  • CUDA Toolkit: 11.7+ (11.2+ for library build and run)

  • GCC: 11.0+ (9.0 for library build and run)

  • Python: 3.7, 3.8, 3.10

Refer to documentation of the sample applications for dependencies.

Known Issues/Limitations

  • Open compilation issue with CUDA Toolkit 11.2 + GCC 10.3

License

CV-CUDA operates under the Apache 2.0 license.

Resources

  1. CV-CUDA GitHub

  2. CV-CUDA Corporate Blog Announcement

  3. CV-CUDA High Performance Image Process Acceleration Library TechBlog

Acknowledgements

CV-CUDA is developed jointly by NVIDIA and the ByteDance Machine Learning team.