this code implements the method proposed in paper "Attention Graph Convolution Network for Image Segmentation in Big SAR Imagery Data". if it helps you, please kindly cite this paper. https://doi.org/10.3390/rs11212586

1. run "python generate_gt.py ", to generate the ground truth data from GT.image
2. run "python main.py" to train and test our model,and print the pixel-level Evaluation Metrics

train, test our AGCN, and calculate the pixel-level Evaluation Metrics（Kappa, precison, recall, and confusion matrix）. Note: as comparison, this file also provide the code of GCN and GAT. If you want to see the results of GAT or GCN, set two parameters "model_nm" and "model" as GAT or GCN

generate the ground truth for training the Network

define the layers

the trained feature_extraction_net

define graph convolution Network

train the feature_extraction_net

define the function for calculating pixel-level Evaluation Metrics

the code runs in python 2.7

any problem please email me : [email protected]