Skip to content

Latest commit

 

History

History
200 lines (145 loc) · 7.51 KB

README.md

File metadata and controls

200 lines (145 loc) · 7.51 KB

Forks Stargazers Issues Python MIT License LinkedIn


Logo

Semantic Segmentation

This is a repository to train semantic segmetation models. It will be improved over time.
Explore the docs »

View Demo · Report Bug · Request Feature

Table of Contents
  1. About The Project
  2. Results
  3. Getting Started
  4. Usage
  5. Roadmap
  6. Contributing
  7. License
  8. Contact
  9. Acknowledgements

About The Project

This project started as my master thesis. I will keep improving it as long as i can.

Built Using

  • Python 3.7
  • PyTorch 1.2.0
  • torchvision 0.4.0

Results

Backbone Head Loss mixup mixwh mIoU Weights
resnet101 ACE2P scp 0 0 63.98 drive
resnet101 Deeplabv3+ CE 0 0 62.86 drive
resnet101 Deeplabv3+ CE 1 1 66.69 drive
resnet101v2 Deeplabv3+ CE 0 0 68.34 drive
resnet101v2 Deeplabv3+ RMI 0 0 69.34 drive
resnet101v2 Deeplabv3+ RMI 1 0 69.38 drive
resnet101v2 Deeplabv3+ RMI 1 1 69.53 drive
resnet101v2 Deeplabv3+ RMI+EL 1 1 69.6 drive

Getting Started

To get a local copy up and running follow these simple steps.

Requierments

First install the requierments as followed.

pip install -r requirments.txt

Installation

  1. Clone the repo 
    git clone https://github.com/MR3z4/SemanticSegmentation.git
  2. Run the training code 
    python main.py

Usage

It will be completed over time.

Roadmap

  • Multi GPU support
  • Add RMI loss
  • Add Mixup option for training
  • Add Mixup Without Hesitation for training with mixup
  • Add AdaBelief optimizer option for training
  • Add CE2P Network(with normal BatchNorm) for training.
  • Add InPlace Active BatchNorm for CE2P.
  • Add SCHP completely.
  • Add Edge Loss
  • Add Vision Transformers
  • Add MixMatch option for training
  • Add FixMatch option for training
  • Distibuted support

See the open issues for a list of proposed features (and known issues).

Contributing

Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.

  1. Fork the Project
  2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
  3. Commit your Changes (git commit -m 'Add some AmazingFeature')
  4. Push to the Branch (git push origin feature/AmazingFeature)
  5. Open a Pull Request

License

Distributed under the MIT License. See LICENSE for more information.

Contact

M.Mohammadzade - @realMReza - mohammadzade.m.r@gmail.com

Project Link: https://github.com/MR3z4/SemanticSegmentation

Acknowledgements

  • Peike Li, Yunqiu Xu, Yunchao Wei, Yi Yang. "Self-Correction for Human Parsing" IEEE Transactions on Pattern Analysis and Machine Intelligence 2020, arXiv:1910.09777, Project Code
  • Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, David Lopez-Paz. "mixup: Beyond Empirical Risk Minimization." arXiv preprint arXiv:1710.09412. arXiv:1710.09412
  • Hao Yu, Huanyu Wang, Jianxin Wu. "Mixup Without Hesitation" arXiv preprint arXiv:2101.04342. arXiv:2101.04342
  • Shuai Zhao, Yang Wang, Zheng Yang, Deng Cai. "Region Mutual Information Loss for Semantic Segmentation", NeurIPS 2019, arXiv:1910.12037, Project Code
  • Juntang Zhuang, Tommy Tang, Yifan Ding , Sekhar Tatikonda, Nicha Dvornek, Xenophon Papademetris, James S. Duncan. "AdaBelief Optimizer: fast as Adam, generalizes as good as SGD, and sufficiently stable to train GANs." arXiv:2010.07468, Project Code