by Quande Liu, Cheng Chen, Jing Qin, Qi Dou, Pheng-Ann Heng.

This repository is for our CVPR 2021 paper 'FedDG: Federated Domain Generalization on Medical Image Segmentation via Episodic Learning in Continuous Frequency Space'.

1. Start with a demo for continuous frequency space interpolation among federated clicnets:

   python freq_space_interpolation_demo.py

2. Prepare the dataset, and then extract the amplitude spectrum of samples in each local client with the function in dataset/prepare_dataset.py:

3. Organize the data (save the data as npy to speed up federated training) and amplitude spectrum of local clients as following structure:

     ├── dataset
        ├── client1
           ├── data_npy
               ├── sample1.npy, sample2.npy, xxxx
           ├── freq_amp_npy
               ├── amp_sample1.npy, amp_sample2.npy, xxxx
        ├── clientxxx
        ├── clientxxx
   

4. Train the federated learning model with ELCFS:

   python train_ELCFS.py

If this repository is useful for your research, please consider citing:

@article{liu2021feddg,
  title={FedDG: Federated Domain Generalization on Medical Image Segmentation via Episodic Learning in Continuous Frequency Space},
  author={Liu, Quande and Chen, Cheng and Qin, Jing and Dou, Qi and Heng, Pheng-Ann},
  journal={The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2021}
}

Some of the code is adapted from SAML and FDA. The datasets used in this paper are downloaded from Prostate and Fundus.

Please contact 'qdliu0226@gmail.com'