Pytorch implementation of MICCAI'23 paper Source-Free Domain Adaptive Fundus Image Segmentation with Class-Balanced Mean Teacher.
- Install python 3.10.5, pytorch 1.12.0, CUDA 11.6 and other essential packages (Note that using other versions of packages may affect performance.)
- Clone this repo
git clone https://github.com/lloongx/SFDA-CBMT.git
cd SFDA-CBMT
- Download datasets from here.
- Download source domain model from here or specify the
--data-dir
in./train_source.py
and then run it. - Save source domain model into folder
./logs_train/
. - Run
./train_target.py
with specified--model-file
and--data-dir
to start the SFDA training process.
This repo benefits from BEAL and SFDA-DPL. Thanks for their wonderful works.