Use your triplet loss to finetune arcface recognition model

[azadehashouri]

Hello dear Alfonmedela.

As far as I remember, I had asked you question many years ago in another repository. :))

Now, I'm working on finetuning Arcface-torch recognition model. My custom dataset has 2 images per identity and the number of total identities is about 4000.

Do you think I can use the triplet loss to finetune it? And how?

Thanks in advance