A method is based on a U-Net architecture and integrates two attention mechanisms: channel attention of squeeze-and-excitation (SE) blocks with residual connections, spatial attention of pyramid pooling module (PPM).
another implement via pytorch: https://github.com/benzzhang/DARU-Net-Pytorch
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|