krumo
commented
3 years ago Sorry for the late reply.
The loss_classifier, loss_box_reg, loss_objectness, loss_rpn_box_reg are losses defined in the paper Faster R-CNN, which corresponds to the classification loss of bbox head, bbox regression loss of bbox head, binary classification loss of rpn, bbox regression loss of rpn.
The loss_da_image, loss_da_instance, loss_da_consistency are losses defined in the paper Domain Adaptive Faster R-CNN, which corresponds to the loss for image-level domain adaptation, loss for instance-level domain adaptation and the consistency regularization for aligning the prediction of image-level domain discriminator and instance-level domain discriminator.
I would refer you to the original papers to find a more detailed explanation about the losses mentioned above.
Hello. I checked your new git. Thanks for your great work.
My question is the meaning of many losses.
where can I found detailed explanation of these loss?