Code for the MICCAI 2022 (early accepted) paper: "Toward Clinically Assisted Colorectal Polyp Recognition via Structured Cross-modal Representation Consistency"

We only provide the public part [link] with its bbox we labeled [link] of the CPC-paired Dataset while the another part is private. Thanks for understanding. And you should label the bounding box of the polyp and crop for preprocessing.

pip install -r requirements.txt

We provide the pretrained weights of the model which were mentioned in the paper.

• Base Weights (vit_small_patch16_224_in21k) link

For other model sizes, please refer to this table and download on the Internet or contact us to provide.

• Smaller Weights (vit_tiny_patch16_224_in21k)
• Larger Weights (vit_base_patch16_224_in21k)

we set the argument "fold" (default: 0) for k-fold cross validation, you can omit it if unnecessary.

python main.py --data_path /the/data/path/ --weights /the/pretrained/weights/path/

If you use any part of this code and pretrained weights for your own purpose, please cite our paper.

@InProceedings{Ma-CPC-Trans,
  author={Ma, Weijie and Zhu, Ye and Zhang, Ruimao and Yang, Jie and Hu, Yiwen and Li, Zhen and Xiang, Li},
  title={Toward Clinically Assisted Colorectal Polyp Recognition via Structured Cross-Modal Representation Consistency},
  booktitle={Medical Image Computing and Computer Assisted Intervention -- MICCAI 2022},
  year={2022},
  publisher={Springer Nature Switzerland},
  pages={141--150}
}

• Weijie Ma