super-wcg

Results 5 issues of super-wcg

I don't understand how to update the self.softmax() when training, because the acc don't backward. How does it work?

@jefflai108 Could you provide your spk2idx file when train the spk_class.py?

When I train the model used your code, I can not achieve 11.7%. Could you show the details?

Because there are overlap bboxes in the same object, when compute the mAP, whether the overlap bboxes influence the mAP or not?

I run this code in Caltech with paper's setting,but the best result tested by dbEval.m is 9.8. Anyone knows ?