Add new distance losses - Githubissues

isaaccorley / torchseg

Segmentation models with pretrained backbones. PyTorch.

MIT License

104 stars 8 forks source link

Add new distance losses #10

Open notprime opened 9 months ago

notprime commented 9 months ago

We should add these two losses:

original distance loss from U-Net
Boundary Loss, used in highly unbalanced segmentation scenarios - GitHub