Awesome

U-Net Semantic Segmentation For Remote Sensing Images


image	label	predict

This code is mainly to solve binary classification semantic segmentation

Here an example is given by using Inria Aerial Image Labeling Dataset. and

/train/ - this folder contains the training set images

/train/ image/ - the folder contains the images who have been cut to a specific size from remote sensing images

/train/ label/ - the folder contains the labels corresponding to the /train/ image/
/val/ - this folder contains the validation set images consistent with the training set structure
/test/ - this folder contains the test set images consistent with the training set structure

Directly run train.py functions with different network parameter settings to produce the results.

test.py can predict images in test set and save them, after that iou.py can calculate oa, F1 score Etc. on test set

predict_rsimage.py can predict a single large Remote sensing image and save it

split.py can split Remote sensing images to specific size for building dataset structure

I have used utility functions from other wonderful open-source projects. Espeicially thank the authors of: