- [Get pre-trained model in this link] (https://drive.google.com/drive/folders/1UC3XOoezeum0uck4KBVGa8osahs6rKUY?usp=sharing): Put pretrained Swin-T into folder "pretrained_ckpt/" and create dir 'chechpoint','test_log' in the root path.
- You can also go to https://challenge.isic-archive.com/data/#2017 to acquire the ISIC2017 dataset. , and then fill in the corresponding path in
./datasets/process.pybefore running it. Change the imgs_train_path, imgs_val_path, imgs_test_path in the train_ISIC to the path of the corresponding path. - You can also go to https://www.kaggle.com/datasets/cf77495622971312010dd5934ee91f07ccbcfdea8e2f7778977ea8485c1914df to acquire the COVID-QU-Ex dataset. See
./datasets/train_covid.txt,./datasets/val_covid.txt,./datasets/test_covid.txtfor the way the data is divided for training, validation, and testing.
- Please prepare an environment with python=3.8, and then use the command "pip install -r requirements.txt" for the dependencies.
-
Run the train script on the ISIC-2017 and the COVID-QU-Ex dataset. The batch size we used is 8 and 16. If you do not have enough GPU memory, the bacth size can be reduced to 6 or 12 to save memory. For more information, contact [email protected].
-
Train
python train_ISIC.py
python train_COV.py
- Test
python test_ISIC.py
python test_COV.py
If you find the codebase useful for your research, please cite the papers:
@article{chen2023cotrfuse,
title={CoTrFuse: a novel framework by fusing CNN and transformer for medical image segmentation},
author={Chen, Yuanbin and Wang, Tao and Tang, Hui and Zhao, Longxuan and Zhang, Xinlin and Tan, Tao and Gao, Qinquan and Du, Min and Tong, Tong},
journal={Physics in Medicine \& Biology},
volume={68},
number={17},
pages={175027},
year={2023},
publisher={IOP Publishing}
}