This is an official release of the paper Dual Multi-scale Mean Teacher Network for Semi-supervised Infection Segmentation in Chest CT Volume for COVID-19.

Dual Multi-scale Mean Teacher Network for Semi-supervised Infection Segmentation in Chest CT Volume for COVID-19,
> Liansheng Wang, Jiacheng Wang, Lei Zhu, Huazhu Fu, Ping Li, Gary Cheng, Zhipeng Feng, Shuo Li, and Pheng-Ann Heng
In: IEEE Transactions on Cybernetics (T-CYB), 2023
[arXiv][Bibetex]

• [2/6 2023] We have uploaded the training codes.
• [11/10 2022] This paper has been accepted to T-CYB.
• [10/25 2022] We have created the repo.

• Network
• Pre-processing
• Pre-trained Weights
• Test codes
• Training Codes

For more details or any questions, please feel easy to contact us by email (jiachengw@stu.xmu.edu.cn).

Please download the dataset of COVID-19-P20 and MOSMED.

The file contains the pre-processing tools for both datasets. Please replace the data path with yours and then run,

$ python scripts/prepare_data.py

Before semi-training the network, you could train the basic parameters under full-supervision for the soft initialization, i.e.,

$ python scripts/train.py $PARAM

Then, you could refine the parameters using extensive unlabeled data, i.e.,

$ python scripts/semi-train.py $PARAM

Please change the $PARAM to your desired inputs.

You could download the pre-trained weights from BaiDu Disk (g2hd). Please store it locally with correct path, i.e., logs/mosmed/dmmtnet_multi_mt_0.1. Then, please run,

$ python scripts/test.py --gpu 0 --arch dmmtnet --dataset mosmeed

If you find DM2TNet useful in your research, please consider citing:

@article{wang2022dual,
  title={Dual Multiscale Mean Teacher Network for Semi-Supervised Infection Segmentation in Chest CT Volume for COVID-19},
  author={Wang, Liansheng and Wang, Jiacheng and Zhu, Lei and Fu, Huazhu and Li, Ping and Cheng, Gary and Feng, Zhipeng and Li, Shuo and Heng, Pheng-Ann},
  journal={IEEE Transactions on Cybernetics},
  year={2022},
  publisher={IEEE}
}