Home

Awesome

Semi-supervised Spatial Temporal Attention Network for Video Polyp Segmentation

Introduction

This repository contains the fully-supervised(fully-supervised for training, unsupervised for testing) version in SUN-SEG dataset of:

Semi-supervised Spatial Temporal Attention Network for Video Polyp Segmentation, MICCAI 2022 PDF

Although our work was based on semi-supervised learning, the fully-supervised version still achieves state-of-the-art results in most metrics.

Usage

This repository is based on GewelsJI/VPS, we strongly recommend you read their work first.

Results

You can directly download the prediction maps of our approach in this download link: MEGA

Existing Methods: source

<p align="center"> <img src="./assets/ModelPerformance.png"/> <br /> </p>

Ours:

DatasetMethodSmeasuremeanEmwFmeasuremeanFmmaxDicemeanSen
SUN-SEG-Easy(Unseen)2022-MICCAI-SSTAN0.8050.8380.6910.7450.7260.662
SUN-SEG-Hard(Unseen)2022-MICCAI-SSTAN0.8010.8330.6820.7340.7180.676

Citations

If you feel this work is helpful, please cite our paper

@inproceedings{zhao2022semi,
  title={Semi-supervised Spatial Temporal Attention Network for Video Polyp Segmentation},
  author={Zhao, Xinkai and Wu, Zhenhua and Tan, Shuangyi and Fan, De-Jun and Li, Zhen and Wan, Xiang and Li, Guanbin},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={456--466},
  year={2022},
  organization={Springer}
}

Changes

Acknowledgements