Hi,
I wanted to test other loss function for Mask R-CNN, so I followed this answer here.
This is what I did as a test:
- I took
maskrcnn_loss
, changed the name, and added aprint
to make sure that everything was ok. - I tried to use
roi_heads.mask_rcnn_loss = My_Loss
- And I alsoI tried to use
mymodel.roi_heads.mask_rcnn_loss = My_Loss
Unfortunately, in both case, MyLoss
was never called (print never executed).
I use TorchVision 0.6.1.
What am I doing wrong?