search

YirongMao / softmax_variants

PyTorch code for softmax variants: center loss, cosface loss, large-margin gaussian mixture, COCOLoss, ring loss
252 stars 49 forks source link

Margin in LMCL loss might have some problems. #1

Closed ZhangXLaurence closed 6 years ago

ZhangXLaurence commented 6 years ago

Hi, According to LMCL paper, the margin was only computed between feature x and its own class weight W. Hence I think label should be passed into your implementation and LMCL_loss should be modified.

YirongMao commented 6 years ago

Hello, The code for LMCL paper is implemented by myself, not by the original authors. I didn't test my code on face datasets. If you test it and find any problem, be free to contact me.