Dual_Adaptive_Graph_Reasoning

• please refer to the implementation code: https://doi.org/10.5281/zenodo.6539438

• Prepare your data.

• Download the pre-train Res2Net parameters and put it into ./lib/res_weight/

• https://drive.google.com/file/d/16_bh91WdaAKPFnwkMkDNMaQUTTIHboWi/view?usp=sharing

Train

• run python train.py

Test

• Download our pre-trained model's parameter then put it into your-own-path
• https://drive.google.com/file/d/12uqe20GdX3-DxeszPqdlIB2o8svyh_vs/view?usp=sharing
• run test.py

Qualitative Results of Coherence Comparison

Citation

If you find our work useful or our work gives you any insights, please cite:

@article{meng2022dual,
  title={Dual Consistency Enabled Weakly and Semi-Supervised Optic Disc and Cup Segmentation with Dual Adaptive Graph Convolutional Networks},
  author={Meng, Yanda and Zhang, Hongrun and Zhao, Yitian and Gao, Dongxu and Hamill, Barbra and Patri, Godhuli and Peto, Tunde and Madhusudhan, Savita and Zheng, Yalin},
  journal={IEEE Transactions on Medical Imaging},
  year={2022},
  publisher={IEEE}
}