VAFF-Net

This repository holds the Pytorch implementation of VAFF-Net.

Introduction

We propose a Voting-based Adaptive Feature Fusing Multi-task Network (VAFF-Net) for joint learning of retinal vessel (RV), foveal avascular zone (FAZ), and retinal vascular junction (RVJ) in OCTA images. In addition, our proposed method can be used as a general multi-task learning framework, and We validate it on the public DRIVE dataset.

Getting Started

Clone this repo

git clone https://github.com/iMED-Lab/VAFF-Net.git

Install prerequisites

cd VAFF-Net
pip install -r requirements.txt

Prepare your data

Please put the root directory of your dataset into the folder ./data. The root directory contain the two subfolder now: ROSE-MT (the public OCTA dataset with multi-task annotations), DRIVE-MT (the public fundus dataset with multi-task annotations).

You can change the path of the dataset and other configurations in the ./config.py

The information about the ROSE dataset with multi-task annotations could be seen in the following link:

https://imed.nimte.ac.cn/ROSE-O.html

Running the code

Start Visdom

    python -m visdom.server  -p 2333

Training on the OCTA dataset

    python train-OCTA.py

Training on the DRIVR dataset

    python train-DRIVE.py

Citation

If you use this code for your research, please cite our papers.

@article{hao2022vaffnet,
  author={Hao, Jinkui and Shen, Ting and Zhu, Xueli and Liu, Yonghuai and Behera, Ardhendu and Zhang, Dan and Chen, Bang and Liu, Jiang and Zhang, Jiong and Zhao, Yitian},
  journal={IEEE Transactions on Medical Imaging}, 
  title={Retinal Structure Detection in OCTA Image via Voting-based Multi-task Learning}, 
  year={2022},
  pages={1-1},
  doi={10.1109/TMI.2022.3202183},
}

iMED-Lab / VAFF-Net