CV-CUDA
CV-CUDA™ is an open-source, GPU accelerated library for cloud-scale image processing and computer vision.
What is CV-CUDA?
CV-CUDA™ is an open-source, GPU accelerated library for cloud-scale image processing and computer vision.
Overview of CV-CUDA
What it is
CV-CUDA is an open-source library of GPU-accelerated computer vision algorithms designed for speed and scalability. It delivers high-throughput, low-latency image/video processing for AI pipelines across NVIDIA cloud, desktop, and edge platforms. CV-CUDA is built for performance and works seamlessly with C/C++ and Python Image and AI frameworks.
Key features
- GPU-accelerated computer vision algorithms for high-throughput, low-latency processing.
- Supports C/C++ and Python bindings, with Python support for versions 3.10 through 3.14, including free-threaded Python 3.14 (3.14t).
- Available for Linux x86_64 and aarch64 SBSA platforms, with Jetson support varying by platform.
- Compatible with CUDA 12 (≥12.2) and CUDA 13 (≥13.0), and NVIDIA drivers ≥r525 (CUDA 12) or ≥r580 (CUDA 13).
- Supports GPU architectures from Turing, Ampere, Ada Lovelace, Hopper, Blackwell, and Jetson Orin/Thor.
- Can be installed via pre-built packages (Python wheels, Debian packages, or tar archives) or built from source.
Use cases
CV-CUDA is designed for cloud-scale image processing and computer vision tasks. It can be used to accelerate AI pipelines that require high-throughput image resizing, decoding, and other computer vision operations. Example use cases include optimizing visual search efficiency, short-form video processing, and video enhancement AI for PaaS platforms.
Getting started
pip install cvcuda-cu12 # for CUDA 12pip install cvcuda-cu13 # for CUDA 13For complete installation instructions including building from source, installing Debian packages, and tar archives, see the repository README.
License and pricing
CV-CUDA is open-source and operates under the Apache-2.0 license. The pricing type is open_source.
Based on the project's public README and repository metadata.