📄 TriLA: Triple-Level Alignment based Unsupervised Domain Adaptation for Joint Segmentation of Optic Disc and Optic Cup

🍺🍺🍺Cheers! Our TriLA won the "Winner Finalist Award" on the task3 of MICCAI FLARE 2024. The detailed solution can be found in another repository.

---

This is the official pytorch implementation of our IEEE JBHI 2024 paper "TriLA: Triple-Level Alignment based Unsupervised Domain Adaptation for Joint Segmentation of Optic Disc and Optic Cup". In this paper, we propose a triple-level alignment based unsupervised domain adaptation method (TriLA) to achieve complete domain alignment.

Requirements

Python 3.7
Pytorch 1.8.0

Usage

Installation

• Clone this repo

git clone https://github.com/Chen-Ziyang/TriLA.git
cd TriLA/TriLA-master

Data Preparation

RIGA+ Dataset

How to Run

We take the scenario using BinRushed (source domain) and Base1 (target domain) as the example.

# Training
CUDA_VISIBLE_DEVICES=0 python main.py --mode train_DA --dataset_root YOUR_ROOT --Target_Dataset Base1 --Source_Dataset BinRushed \
--vae_coef 0.1 --output_coef 0.1 --style_coef 0.1 --content_coef 0.001
# Test
CUDA_VISIBLE_DEVICES=0 python main.py --mode test --reload EPOCH_OF_MODEL --load_time TIME_OF_MODEL --Target_Dataset BinRushed

Acknowledgement

Part of the code is revised from the Pytorch implementation of DoCR.

Citation ✏️ 📄

If you find this repo useful for your research, please consider citing the paper as follows:

@article{chen2024trila,
  title={TriLA: Triple-Level Alignment based Unsupervised Domain Adaptation for Joint Segmentation of Optic Disc and Optic Cup},
  author={Chen, Ziyang and Pan, Yongsheng and Ye, Yiwen and Wang, Zhiyong and Xia, Yong},
  journal={IEEE Journal of Biomedical and Health Informatics},
  year={2024},
  volume={28},
  number={9},
  pages={5497--5508},
  publisher={IEEE}
}