search

kuangliu / pytorch-retinanet

RetinaNet in PyTorch
992 stars 250 forks source link

classification branch #48

Open foralliance opened 5 years ago

foralliance commented 5 years ago

@kuangliu HI

For focal loss. The classification branch uses the sigmoid function.

Why background class is considered in the classification branch?? such as coco, num_classes=80 instead of 81.

pengfeidip commented 5 years ago

I think the classification layer lacks sigmoid operation ..

EvanAlbee commented 4 years ago

Same here.