deep supervision loss

[cugwu]

For the deep supervision implemented in dynamic unet of MONAI (https://docs.monai.io/en/stable/_modules/monai/networks/nets/dynunet.html). Is there a particular loss to apply or I just do the sum of the loss of each of the deep supervision heads?

[yiheng-wang-nv]

Hi @cugwu , the loss used is in:

tutorials/modules/dynunet_pipeline/trainer.py

Line 56 in 608795f

0.5 ** i * self.loss_function.forward(p, targets)

I think you can also try to adjust the weights for different resolutions, not sure if there are other methods.