把 YOLO 系列模型接入 NVIDIA DeepStream SDK 的推理实现，支持 Darknet、PyTorch、Paddle 等多种权重格式，方便在 Jetson 和 GPU 上做实时目标检测部署。
基于 GStreamer 和 TensorRT 的 GPU 加速实时视频与多传感器分析 SDK，可在边缘、本地和云端搭建多路视频 AI 推理流水线。
用 Python 把多路视频流、多个模型和多块 GPU 串起来做实时视频分析，底层靠 CUDA、TensorRT 和 DeepStream 加速，适合搭建高吞吐的视觉推理流水线。
Please see the DeepStream 9.2 release notes at: https://docs.nvidia.com/deepstream/9.2/text/DS Release notes.html. --- Release Artifacts Full SDK packages — bare-metal in
Please see the DeepStream 9.1 release notes at: https://docs.nvidia.com/metropolis/deepstream/9.1/text/DS Release notes.html. --- Note on 9.1.1 assets below: This release
Please see the DeepStream 9.0 release notes at: https://docs.nvidia.com/metropolis/deepstream/9.0/text/DS Release notes.html. Changes in 9.0.1: 1) New DeepStream Skills;
Please see the DeepStream 9.0 release notes at: https://docs.nvidia.com/metropolis/deepstream/9.0/text/DS Release notes.html.
Platform: DeepStream 7.0 JetPack (Jetson): 6.x Release manager: @denisvmedyantsev What's Changed Added pyfunc support in the Watchdog service Docs: https://docs.savant-ai
Platform: DeepStream 7.0 JetPack (Jetson): 6.x Release manager: @denisvmedyantsev What's Changed Fixed a bug in the Watchdog service to support Buffer-NG. (by @abramov-ol