Attention Guided Network for Retinal Image Segmentation (AG-Net)

The code of "Attention Guided Network for Retinal Image Segmentation" in MICCAI 2019.

• The code is based on: Python 2.7 + pytorch 0.4.0.
• You can run <AG_Net_path>/code/test.py for testing any new image directly.
• You can run <AG_Net_path>/code/main.py for training a new model.

Quick usage on your data:

• Put your desired file in "\<AG_Net_path>/data/\<your_file_name>".
• Put the images in "\<AG_Net_path>/data/\<your_file_name>/images".
• Put the labels in "\<AG_Net_path>/data/\<your_file_name>/label".
• Divide data into training and test data, and store the image name in the "train_dict.pkl" file. (We provide a 'train_dict.pkl' sample for DRIVE dataset)
• The "train_dict.pkl" should contains two dictionary: 'train_list' and 'test_list'.

Train your model with:

python main.py --data_path '../data/your_file_name'

Reference

1. S. Zhang, H. Fu, Y. Yan, Y. Zhang, Q. Wu, M. Yang, M. Tan, Y. Xu, "Attention Guided Network for Retinal Image Segmentation," in MICCAI, 2019. [PDF]
2. H. Fu, J. Cheng, Y. Xu, D. W. K. Wong, J. Liu, and X. Cao, “Joint Optic Disc and Cup Segmentation Based on Multi-Label Deep Network and Polar Transformation,” IEEE Trans. Med. Imaging, vol. 37, no. 7, pp. 1597–1605, 2018.