#172 · Primary category: Computer Vision

Savant

computer-vision cuda deep-learning deepstream edge-computing inference-engine instance-segmentation machine-learning nvidia nvidia-deepstream-sdk object-detection opencv peoplenet tensorrt video yolo yolov5-face yolov8 yolov8-face

Python Computer Vision & Video Analytics Framework With Batteries Included

Project last updated:08/28/26

GitHub Stars

849

Forks

83

Contributors

19

License

Apache-2.0

Why we included this project

Teams building real-time video analytics on NVIDIA hardware often end up wrestling with DeepStream's low-level plumbing. Savant wraps that stack in a higher-level Python, C++, and Rust framework, so you can assemble dynamic, fault-tolerant inference pipelines without hand-writing the boilerplate for frame handling, stream management, and GPU resources. It targets production workloads such as object detection, instance segmentation, and people and vehicle tracking on data center GPUs or Jetson edge devices, and it ships complete demos you can run end to end. If your team already lives in the NVIDIA ecosystem and wants a layer that makes pipeline composition feel like wiring modules rather than managing buffers, Savant is worth evaluating. That same codebase can span dense server-side processing and resource-constrained edge inference, which is where it gets especially useful.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category