QAMFace icon indicating copy to clipboard operation
QAMFace copied to clipboard

Published 20 hours ago • verified publisher icon MccreeZhao

Any prebuilt weights available?

Open xsacha opened this issue 5 years ago • 5 comments

I notice you only went to epoch 22, which is still stage 1. Are you still training this?

xsacha avatar Jan 18 '20 10:01 xsacha

Hi, thanks for your question. I wonder what does stage 1 means? The learning rate is decayed in epoch [4, 10, 15,20]. And the model is already converged according to the training curve.

MccreeZhao avatar Jan 25 '20 04:01 MccreeZhao

Ah, quick to train!

xsacha avatar Jan 25 '20 13:01 xsacha

Is it possible to provide that pytorch weight file?

xsacha avatar Jan 28 '20 03:01 xsacha

@MccreeZhao Is this what the training graph should look like at epoch 6? From your version. image

xsacha avatar Feb 15 '20 13:02 xsacha

I was unable to reproduce your results. My best epochs: 12 / 17 LFW: 99.78 / 99.63 CFP_FF: 99.7 / 99.7 AGEDB: 97.75 / 97.95

xsacha avatar Feb 27 '20 03:02 xsacha