Anuj Sharma
Anuj Sharma
Okay, I can see it now where it is updated to background cls-score from the teacher https://github.com/microsoft/SoftTeacher/blob/bef9a256e5c920723280146fc66b82629b3ee9d4/ssod/models/soft_teacher.py#L235
In the final classification loss here https://github.com/microsoft/SoftTeacher/blob/bef9a256e5c920723280146fc66b82629b3ee9d4/ssod/models/soft_teacher.py#L243 it seems loss is: ``` # where `w` is the cls-score from teacher for background proposal total_cls_loss = sum(cls_loss_on_all_fg_proposals) + sum(w * cls_loss_on_bg_proposals)...
Great. Thanks a lot for the quick clarifications !! :+1:
@lliuz Hi, yes, the above equation matches the code correctly.
Hi @tianweiy thank you for the clarification. Makes sense :+1: Cheers, Anuj
Hi @tianweiy I was wondering if you had any results on nuScenes with Second Stage that you could please share? It would be helpful to understand what performance to expect...
Thanks for your quick response! Is the Two-Stage approach not the same as BEV Feature aggregation?