Facenet: how much time does it take you to train a model with 0.919 accuracy on LFW?

Created on 13 Oct 2016  路  2Comments  路  Source: davidsandberg/facenet

picture ZHAIXINGZHAIYUE

Most helpful comment

It depends on a few factors, e.g. how fast your disk is, how fast the triplet selection can run on the CPU and how fast the training on the GPU is. For me it takes roughly 72 hours to train for 500 000 steps with triplet loss and the nn4 model.

picture davidsandberg on 14 Oct 2016
👍3

All 2 comments

It depends on a few factors, e.g. how fast your disk is, how fast the triplet selection can run on the CPU and how fast the training on the GPU is. For me it takes roughly 72 hours to train for 500 000 steps with triplet loss and the nn4 model.

davidsandberg picture davidsandberg on 14 Oct 2016
👍3

@davidsandberg thank you very much.

ZHAIXINGZHAIYUE picture ZHAIXINGZHAIYUE on 15 Oct 2016
Was this page helpful?
0 / 5 - 0 ratings

Related issues

tonybaigang picture tonybaigang  路  3Comments
xvdehao picture xvdehao  路  4Comments
MaartenBloemen picture MaartenBloemen  路  3Comments
RaviRaaja picture RaviRaaja  路  3Comments
kuaikuaikim picture kuaikuaikim  路  3Comments