YOLOX(Megvii-BaseDetection) Deploy DeepStream 💕 💥

This project base on https://github.com/Megvii-BaseDetection/YOLOX and https://zhuanlan.zhihu.com/p/391693130

News

Adapt the new version of YOLOX original pre-training model, FPS > 100 - 2021-12-20

Deploy yolox to deep stream, FPS > 70 - 2021-7-21

System Requirements

cuda 11.0+

TensorRT 8+

OpenCV 4.0+ (build with opencv-contrib module) how to build

OpenMP

DeepStream 6.0+

Installation

Make sure you had install dependencies list above

# clone project and submodule
git clone {this repo}

cd {this repo}/nvdsinfer_custom_impl_yolox/

make

use to parse infer postprocess.

Run

cd {this repo}

deepstream-app -c deepstream_app_config.txt

How to build engine?

https://zhuanlan.zhihu.com/p/391693130

About License

For the 3rd-party module and Deepstream, you need to follow their license

For the part I wrote, you can do anything you want