the converage question about blob_loss about blob_loss HOT 5 CLOSED

hmm it is difficult to tell what is going wrong with the information provided. You could try to raise alpha (the weight for the main component). Does your training converge with your "normal" loss function? What is your "normal" loss?

from blob_loss.

emm, i set alpha is 1 and it still cann't converge. "normal" loss is chose dice and BCE, and its can converge, it make me emo....

from blob_loss.

If alpha is set to 1 and beta to 0 blob loss is essentially deactivated. If your training does not converge in this state the problem is somewhere else. I am facing a bit of trouble understanding your problem description. What do you want to demonstrate with the code snippet above? Most of it seems to be copy paste from the code here? :)

from blob_loss.

The code I am using now is the code posted above, and I have modified it for my problem. I don't know what the problem is, very emo. I'll keep trying, thanks

from blob_loss.

closed until more information is provided.

from blob_loss.