search

LIVIAETS / boundary-loss

Official code for "Boundary loss for highly unbalanced segmentation", runner-up for best paper award at MIDL 2019. Extended version in MedIA, volume 67, January 2021.
https://doi.org/10.1016/j.media.2020.101851
MIT License
654 stars 98 forks source link

Need to calculate the background-class loss for Multi-class Segment? #39

Closed hei6775 closed 3 years ago

hei6775 commented 3 years ago

The content of your work is great and gave me a lot of ideas. But I have some questions that I haven't figured out. For multi-class segmentation, such as (ACDC dataset), will your dice loss and bl_loss calculate the four types of losses of 0, 1, 2, 3? Does not calculate the loss of 0 (background) to make the effect better? What kind of thinking are you based on?

HKervadec commented 3 years ago

Hi,

So far, I have experimented multi-class by treating the background (class 0) as yet another class. The rationale actually stems from the binary case.

When there are only two classes, the background is the perfect negative of the foreground; therefore its distance map is simply the reverse of the other one, it doesn't bring much new value so we can skip it.

This property is obviously not true anymore in multi-class (at least in the general case), so I started by supervising the background. Skipping it might still work, but for other reasons (indirrect supervision), and it might vary depending on the task.

Does that answer your question?

Best,

Hoel