Hello,
First of all, thank you for your contribution to the literature. This paper has become an inspiring study for us.
What I want to ask is why you prefer sigmoid focal loss instead of cross entropy loss? If I understand correctly, Mask2Former study is the study that you have benefitted from. In there, cross entropy loss is utilized.
Is there a performance difference between both of them? Did you ablate it?
Hello,
First of all, thank you for your contribution to the literature. This paper has become an inspiring study for us.
What I want to ask is why you prefer sigmoid focal loss instead of cross entropy loss? If I understand correctly, Mask2Former study is the study that you have benefitted from. In there, cross entropy loss is utilized.
Is there a performance difference between both of them? Did you ablate it?