klymya/deepstream-example

deepstream-example

A brief description of how to use deepstream 6.1 with Yolo V5. This repo based on multiple blogposts:

• NVIDIA Deepstream Quickstart
• Getting started with custom NVIDIA Deepstream 6.0 pipelines in Python

Here implemented two pipelines - with and without tiling. Both pipeline process video files and produce resulting videos with detections. Also, the pipeline use different trackers:

• NvDCF - python based pipeline with tiling
• DeepSORT - C++ based pipeline without tiling

Configs

The config files placed in deepstream-example/configs

• Detector: yolov5.txt
• Trackers
  • NvDCF: nvdcf.txt, tracker_config.yml
  • DeepSORT: DeepSORT_config.yml
• Preprocess(tiler): preprocess.txt
• Whole pipeline: pipeline.txt

Steps

0. Clone the repo and init submodules

git clone https://github.com/klymya/deepstream-example.git && cd deepstream-example && git git submodule init

1. Download and convert yolov5 weights. The instruction and script for this are here. Put the conversion artifats(.cfg and .wts files) to deepstream-example/weights.

2. Build the docker image with required environment

docker build . -t deepstream-example

3. Run the docker container

docker run -it --rm --gpus all -v ${PWD}:/app -w /app deepstream-example bash

4. Run examples
   1. Python pipeline with tiling

   export PYTHONPATH=/app/code/boilerplate/deepstream && python3 code/main.py
   

   2. C++ pipeline

   deepstream-app -c configs/pipeline.txt
   

Tiling links:

• https://developer.nvidia.com/blog/applying-inference-over-specific-frame-regions-with-nvidia-deepstream/
• https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_plugin_gst-nvdspreprocess.html

Other useful links:

• https://github.com/marcoslucianops/DeepStream-Yolo
• https://github.com/ml6team/deepstream-python
• https://github.com/NVIDIA-AI-IOT/deepstream_python_apps