Hi, I want to train a robust model with voxceleb2 original data and voxceleb2 augmentation data (might add any data later). Could you give some advice about which is a better way to do it, train from scratch or finetune the voxceleb_trainer baseline model? (I tried to finetune, but get worse result than baseline model)
Hi, I want to train a robust model with voxceleb2 original data and voxceleb2 augmentation data (might add any data later). Could you give some advice about which is a better way to do it, train from scratch or finetune the voxceleb_trainer baseline model? (I tried to finetune, but get worse result than baseline model)