The contrast loss - Githubissues

HarryVolek / PyTorch_Speaker_Verification

PyTorch implementation of "Generalized End-to-End Loss for Speaker Verification" by Wan, Li et al.

BSD 3-Clause "New" or "Revised" License

576 stars 164 forks source link

The contrast loss #70

Closed amadeusuzx closed 4 years ago

amadeusuzx commented 4 years ago

Hi there, Maybe I just didn't see it, but there seems no scripts implementing the contrast loss, which takes into account of the most aggressive pairs in the similar matrix?

amadeusuzx commented 4 years ago

Sorry, I think I understand now -- TI-SV uses softmax loss .