This is the official pytorch implementation of our AAAI 2025 paper "Gradient Alignment Improves Test-Time Adaptation for Medical Image Segmentation".
CUDA 10.1
Python 3.7.0
Pytorch 1.8.0
CuDNN 8.0.5
Our Anaconda environment is also available for download from Google Drive.
Upon decompression, please move czy_pytorch to your_root/anaconda3/envs/. Then the environment can be activated by conda activate czy_pytorch.
The preprocessed data can be downloaded from Google Drive.
Download pre-trained models from Google Drive and drag the folder 'models' into the folder 'GraTa-master'.
You can also train your own models.
Please first modify the root in run.sh, and then run the following command to reproduce the results.
bash run.sh
If this code is helpful for your research, please cite:
@article{chen2025grata,
title={Gradient Alignment Improves Test-Time Adaptation for Medical Image Segmentation},
author={Chen, Ziyang and Ye, Yiwen and Pan, Yongsheng and Xia, Yong},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
year={2025}
}
Ziyang Chen (zychen@mail.nwpu.edu.cn)