When training with voc datset, why is the output channel of the classifier subnet 9*20*H*W instead of 9*21*H*W? #45

Open LpfML opened 5 years ago