RETFound

RETFound is pre-trained on 1.6 million retinal images with self-supervised learning
RETFound has been validated in multiple disease detection tasks
RETFound can be efficiently adapted to customised tasks

🎉News

🐉2025/02: We organised the model weights on HuggingFace, no more manual downloads needed!
🐉2025/02: Multiple pre-trained weights, including MAE-based and DINOV2-based, are added!
🐉2025/02: We update the version of packages, such as CUDA12+ and PyTorch 2.3+!
🐉2024/01: Feature vector notebook are now online!
🐉2024/01: Data split and model checkpoints for public datasets are now online!
🎄2023/12: Colab notebook is now online - free GPU & simple operation!
2023/09: a visualisation demo is added
2023/10: change the hyperparameter of input_size for any image size

🔧Install environment

Create environment with conda:

conda create -n retfound python=3.11.0 -y
conda activate retfound

Install dependencies

conda install pytorch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 pytorch-cuda=12.1 -c pytorch -c nvidia
git clone https://github.com/rmaphoh/RETFound_MAE/
cd RETFound_MAE
pip install -r requirements.txt

🌱Fine-tuning with RETFound weights

To fine tune RETFound on your own data, follow these steps:

Get access to the pre-trained models on HuggingFace (register an account and fill in the form) and go to step 2:

	ViT-Large	Source
RETFound_mae_natureCFP	access	Nature RETFound paper
RETFound_mae_natureOCT	access	Nature RETFound paper
RETFound_mae_meh	access	TBD
RETFound_mae_shanghai	access	TBD
RETFound_dinov2_meh	access	TBD
RETFound_dinov2_shanghai	download	TBD

huggingface-cli login --token YOUR_HUGGINGFACE_TOKEN

Organise your data into this directory structure (Public datasets used in this study can be downloaded here)

├── data folder
    ├──train
        ├──class_a
        ├──class_b
        ├──class_c
    ├──val
        ├──class_a
        ├──class_b
        ├──class_c
    ├──test
        ├──class_a
        ├──class_b
        ├──class_c

Start fine-tuning (use IDRiD as example). A fine-tuned checkpoint will be saved during training. Evaluation will be automatically run after training.

model can be "RETFound_mae" or "RETFound_dinov2"

finetune can be "RETFound_mae_natureOCT", "RETFound_mae_natureCFP", "RETFound_mae_meh", "RETFound_mae_shanghai", "RETFound_dinov2_meh", and "RETFound_dinov2_shanghai".

torchrun --nproc_per_node=1 --master_port=48798 main_finetune.py \
    --model RETFound_mae \
    --savemodel \
    --global_pool \
    --batch_size 16 \
    --world_size 1 \
    --epochs 100 \
    --blr 5e-3 --layer_decay 0.65 \
    --weight_decay 0.05 --drop_path 0.2 \
    --nb_classes 5 \
    --data_path ./IDRiD \
    --input_size 224 \
    --task RETFound_mae_meh-IDRiD \
    --finetune RETFound_mae_meh

For evaluation only (download data and model checkpoints here; change the path below)

torchrun --nproc_per_node=1 --master_port=48798 main_finetune.py \
    --model RETFound_mae \
    --savemodel \
    --eval \
    --global_pool \
    --batch_size 16 \
    --world_size 1 \
    --epochs 100 \
    --blr 5e-3 --layer_decay 0.65 \
    --weight_decay 0.05 --drop_path 0.2 \
    --nb_classes 5 \
    --data_path ./IDRiD \
    --input_size 224 \
    --task RETFound_mae_meh-IDRiD \
    --resume ./finetune_IDRiD/checkpoint-best.pth

📃Citation

If you find this repository useful, please consider citing this paper:

TBD

@article{zhou2023foundation,
  title={A foundation model for generalizable disease detection from retinal images},
  author={Zhou, Yukun and Chia, Mark A and Wagner, Siegfried K and Ayhan, Murat S and Williamson, Dominic J and Struyven, Robbert R and Liu, Timing and Xu, Moucheng and Lozano, Mateo G and Woodward-Court, Peter and others},
  journal={Nature},
  volume={622},
  number={7981},
  pages={156--163},
  year={2023},
  publisher={Nature Publishing Group UK London}
}