Semi-supervised Pytorch Medical Segmentation

Recent Updates

• 2022.5.11 The SSL codes are released.

Requirements

• PyTorch > 1.7
• torchio <= 0.18.20
• python >= 3.7

Notice

• You can modify hparam.py to determine whether 2D or 3D segmentation.
• We provide some SSL algorithms on 2D and 3D segmentation.
• The project is highly based on Pytorch-Medical-Segmentation.

Done

• SSL

• Mean-Teacher
• Cross-Pesudo-Supervision
• Entropy-Minimization
• Interpolation-Consistency
• Co-Training
• Cross-Teaching
• Uncertain-Aware-Mean-Teacher
• Pi-Model

By The Way

This project is not perfect and there are still many problems. If you are using this project and would like to give the author some feedbacks, you can send Me an email.

Acknowledgements

This repository is an unoffical PyTorch implementation of Semi-Supervised Learning Medical segmentation in 3D and 2D and highly based on Pytorch-Medical-Segmentation and SSL4MIS. Thank you for the above repo. The project is done with the supervisions of Prof. Ruoxiu Xiao and Dr. Cheng Chen.