CVCUDA/CV-CUDA

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

2729 stars · 264 forks observed · 2026-10-11

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

CV-CUDA is an open-source GPU-accelerated library of computer vision and image processing operators built on CUDA, with C++ and Python APIs. It is designed for high-throughput, low-latency image and video preprocessing in AI inference pipelines on NVIDIA cloud, desktop, and edge platforms.

Use cases

• resize and preprocess images on GPU before feeding a deep learning model
• batch decode and transform JPEG images for cloud-scale inference
• accelerate video frame preprocessing in AI pipelines
• replace CPU OpenCV preprocessing with GPU-accelerated equivalents
• build end-to-end GPU image processing pipelines in Python or C++

When to choose

• you need high-throughput image preprocessing on NVIDIA GPUs for ML inference
• you want a C++ or Python API that integrates with TensorRT, PyTorch, or other AI frameworks
• you are processing large batches of images or video at cloud scale

When to avoid

• you need CPU-only image processing without NVIDIA GPUs
• you need a general-purpose image editor or GUI tool
• your target platform lacks CUDA support

Facets

library · maturity active

image-processing computer-vision gpu-computing machine-learning computer-vision image-processing gpu-computing deep-learning windows python cpp cross-platform cuda nvidia image-decoding preprocessing batch-processing video-processing linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="CVCUDA/CV-CUDA")