nilskk

So it only multiplies the objectness score with the classification score during inference, because I also got the impression reading the paper that the multiplication would happen during the training...

I would also like to know the answer :)

Did you found any solutions to this?