Closed Royzon closed 3 years ago
I'm curious about your IDloss looks like different from the original paper which trained with triplet loss using the BatchHard strategy for sampling triplets.
i am sorry,it's just a quick implementation of the network structure,training strategy and loss i didn't refer the paper.
ok, thanks
I'm curious about your IDloss looks like different from the original paper which trained with triplet loss using the BatchHard strategy for sampling triplets.