Deploy Lite_HRNet with ONNX format

Overview

The base model originated from Lite-HRNet. In this project, I converted the base model to convert it into an ONNX model for cross-platform compatibility. This model can then be deployed locally from scratch, or by using Docker image.

Check section Tasks for more details

Get started

Build app from scratch

1. (Optional) Create virtual environment

   python -m venv venv
   source venv/bin/activate
   

2. Install dependencies for FastAPI app

   pip install -r requirements.txt
   

3. (Optional) Install dependencies for all files in the project

   pip install -r requirements_full.txt
   

4. Use Demo.ipynb for an interactive demo.
5. Run the app

   uvicorn main:app --reload
   

Build app from Docker image

1. Pull Image from Docker Hub

   docker pull tiviluson/litehrnet_onnx:v0.3
   

2. Run container from the previously pulled image

   docker run -d -p 8000:8000 tiviluson/litehrnet_onnx:v0.3
   

Test the app

1. Run test.py to test the server. Use python -m test.py --help for more information.

Tasks

• Run inference with model Lite_HRNet with MMPose's API
• Rewrite the model to convert it to ONNX format
• Run inference with model Lite_HRNet in ONNX format
• Visualize output
• Build FastAPI server to demostrate the project
• Build Dockerfile for the project
• Run and test Dockerfile for the project
• Deploy the model using Docker and Nginx on AWS EC2
• Deploy the model using K8S with ArgoCD, following GitOps principle. K8S Specs can be found in this repo

Citation

@inproceedings{Yulitehrnet21,
  title={Lite-HRNet: A Lightweight High-Resolution Network},
  author={Yu, Changqian and Xiao, Bin and Gao, Changxin and Yuan, Lu and Zhang, Lei and Sang, Nong and Wang, Jingdong},
  booktitle={CVPR},
  year={2021}
}

@inproceedings{SunXLW19,
  title={Deep High-Resolution Representation Learning for Human Pose Estimation},
  author={Ke Sun and Bin Xiao and Dong Liu and Jingdong Wang},
  booktitle={CVPR},
  year={2019}
}

@article{WangSCJDZLMTWLX19,
  title={Deep High-Resolution Representation Learning for Visual Recognition},
  author={Jingdong Wang and Ke Sun and Tianheng Cheng and 
          Borui Jiang and Chaorui Deng and Yang Zhao and Dong Liu and Yadong Mu and 
          Mingkui Tan and Xinggang Wang and Wenyu Liu and Bin Xiao},
  journal={TPAMI}
  year={2019}
}