FocusNet: Transformer-enhanced Polyp Segmentation with Local and Pooling Attention

April 26, 2025 ยท View on GitHub

Overview

Create Environment

conda create -n FocusNet python==3.8.16
conda activate FocusNet

Install Dependencies

pip install -r requirements.txt

Download Checkpoint

Download pretrained checkpoints from Google Drive and move it to the pretrained_pth directory.

Download Dataset

Download PolypDB dataset from this link and move it to the data directory.

Train

python train_modality_wise.py 

Test

python test_modality_wise.py 

Citation

Please cite our paper if you find the work useful:

@article{zeng2025focusnet,
  title={FocusNet: Transformer-enhanced Polyp Segmentation with Local and Pooling Attention},
  author={Zeng, Jun and Santosh, KC and Nayak, Deepak Rajan and de Lange, Thomas and Varkey, Jonas and Berzin, Tyler and Jha, Debesh},
  journal={arXiv preprint arXiv:2504.13597},
  year={2025}
}

Contact

Please contact zeng.cqupt@gamil.com for any further questions.