Loss and Acc on VGGVox (Thin-Resnet & NetVLAD)

a-nagrani / VGGVox

VGGVox models for Speaker Identification and Verification trained on the VoxCeleb (1 & 2) datasets

377 stars 85 forks source link

Loss and Acc on VGGVox (Thin-Resnet & NetVLAD) #19

Open jjeremy40 opened 3 years ago

jjeremy40 commented 3 years ago

Hi, correct me if I'm wrong but you trained VGGVox on VoxCeleb2 as Classification, right ? If so, what kind of Loss and Acc did you get ? for training set and validation set ? Thanks.