Gated Shape Unet for Semantic Segmentation Inspired from GSCNN
Torch: https://pytorch.org/
pip install matplotlib tqdm sklearn
sudo pip3 install opencv-python
sudo pip3 install scikit-learn
CUDA : https://pytorch.org/get-started/locally/
ISPRS Vaihingen semantic segmentation dataset download link
Change the root dir of the dataset in network.dataset.vaihingen
No data augmentation is done here, it is up to you to implement it (albumentations should work)
Ground truth, 30th epoch segmentation, 42th sepoch segmentation, 42th epoch edges
Confusion matrices of 30th and 42th epochs
