Closed djskwh closed 1 year ago
Hi, @djskwh.
You can just define the whole segmentation model as the base
and torch.nn.Identity()
as the classifier
, which should work properly. But you still need to make your custom model class inherited from DecoupledModel
.
If not, sorry I haven't try to solve any segmentation tasks and can't give you more idea. 😂
Thank you for your advice. Actually, that answer was what i looking for!!
Hi KarhouTam.
I'd like to implement your fed benchmark setting to my segmentation task.
but i think the structuring segmentation model properly to your setup is little bit tricky.
what i understand for properly setup to your work, it requires base layer and classifier layer and etc. ....
so is there any advice for setting up segmentation model?