MedicalNet
MedicalNet copied to clipboard
question about loss function
Hi, I'm confused about the loss function in the train phase.The loss function you have used is cross entropy Loss and the tensors inputted are different shape.Maybe something wrong? Meanwhile i found others have encountered the same problem that net perform well in the train phase,it's loss value is small,but dice value in the test phase maybe 0.
Hello! Did you found the solution,I also trapped in a similar problem. The dice loss is always zero when testing it on own dataset.