This is an official implementation of Contour-Aware Equipotential Learning for Semantic Segmentation (accepted by IEEE TMM 2022) You can download all models in our paper at Google Drive
Our work (checkpoints) is an add-on approach and can be plugged into existing deep semantic segmentation models.
We achieved below PyTorch implementations to test our method:
- PSPNet & PSANet (including Pascal Voc and Cityscapes)
- (Pascal Voc only) DeepLab V3+
- (Cityscapes only) DeepLab V3+
The first repository is released by the author of PSPNet and PSANet, and we achieved powerful mIoU performances on both datasets with DeepLab V3+.