search

bethgelab / siamese-mask-rcnn

Siamese Mask R-CNN model for one-shot instance segmentation
Other
346 stars 60 forks source link

How to design the loss function？ #23

Closed kidcad1412 closed 4 years ago

kidcad1412 commented 4 years ago

I confuse the loss function,please help.

michaelisc commented 4 years ago

I am not sure I understood your question. Are you asking, what the loss is composed of?

kidcad1412 commented 4 years ago

I mean the loss function is same as the loss function in Mask R-CNN ? How the metric show in the loss function?

michaelisc commented 4 years ago

Exactly. The only difference is that the classification is only done for 2 classes ("is this an object from the same category as the reference or is it background") and not 81 classes ("is this an object from one of 80 predefined categories or is it background"). We thus made no changes to the loss but only to the labels which are conditioned on the reference.

kidcad1412 commented 4 years ago

thank you so much