How to calculate edge loss, dataset has relevent annotation?

[Ying-Kang]

edge feature map down to (w/4, h/4) for dice loss calculating,

however, how can i get the relevent edge groud truth for loss deducting?

it is not mentioned, by edge search or already given by dataset?

[wangskyone]

Here's a way to do it https://github.com/dddb11/MVSS

[SunnyHaze]

You could follow the implementation in this repo: https://github.com/dddb11/MVSS-Net