miracle-center / partially-supervised-multi-organ-segmentation Goto Github PK

Hi, sorry to disturb you again, I meet a problem and hope you can help me~

I defined Task 1, Task 2, Task 3, and Task 4, represents four organs, respectively, and the GT labeled value is 1, 2, 3, 4, respectively.

in calculating loss when y_tags != x_tags:

target_onehot = onehot_transform(target, len(cur_task)+1)

def onehot_transform(tensor, depth):
assert torch.max(tensor) < depth, f"{torch.max(tensor)}/{depth}"

when process Task 2, the corresponding target(GT) label value is 2, that is torch.max(tensor) = 2,

depth = len(cur_task)+1, cur_task = Task 2, so len(cur_task) is equal to 1, so the depth =2

bu if depth =2 and torch.max(tensor)=2, the assert ** will be processed and run stop~~

Did I do something wrong?

Thanks a lot~~waiting for your reply~

Hi, many thanks for your sharing,
Can you provide the running order of your code? the same as nnunet as following?
covert --- nnUNet_plan_and_preprocess ---- nnUNet_train?
Thanks, waiting for your reply.

Hi,

Could you please provide a simple README? It is quite hard to distinguish which part is modified. And may I ask where is the code implementation of the two new loss? Thank you.