facenet-pytorch-glint360k icon indicating copy to clipboard operation
facenet-pytorch-glint360k copied to clipboard

Published 20 hours ago • verified publisher icon tamerthamoqa

embedding vectors dimension

Open Amirzkn opened this issue 2 years ago • 1 comments

Hi @tamerthamoqa ,

Thanks a lot for your great repo. According to FaceNet paper the best dimension for embedded vector is 128. I am curious to know is there any specific reason that you used four time bigger embedded vector dimension to 512?

Amirzkn avatar Apr 13 '22 13:04 Amirzkn

Hello @Amirzkn

Thank you for your kind words. There is no concrete reason, I was trying with size 128, 256, and 512 and have found that size 512 yielded the best performance in the training experiments but not much at all. Keeping the size as the FaceNet paper would be fine.

tamerthamoqa avatar Apr 13 '22 17:04 tamerthamoqa