Mask shapes for dice loss + cross entropy loss

vision
2

Hello Hwarang,

It is correct to combine loss and backpropagate it this way. Your network parameters will be updated with respect to ce_loss and dice.

1 Like