L-GM-loss icon indicating copy to clipboard operation
L-GM-loss copied to clipboard

Questions about Face Verification

Open zorrocai opened this issue 5 years ago • 2 comments

Hi, thanks for your work. I am doing some experiments on LFW dataset for my paper, finding your experiments on Face Verification are interesting but unclear. And I could not reproduce the same result as yours. Could you share your Face verification codes with me? Thank you very much.

zorrocai avatar Jan 05 '19 14:01 zorrocai

Actually, I wonder the architecture of your ResNet-Like 27 layers network. How did you design it? And could the 27-layers network with softmax loss trained on CASIA-WebFace produce the accuracy 98.56 on LFW? I tried 36-layers ResNer-Like network with softmax loss trained on CASIA-WebFace only got 97.85 on LFW...

zorrocai avatar Jan 06 '19 05:01 zorrocai

Hi, do you reproduce the results on face verification?

GuoleiSun avatar Oct 01 '19 09:10 GuoleiSun