sv-rdino trainning data.

modelscope / 3D-Speaker

A Repository for Single- and Multi-modal Speaker Verification, Speaker Recognition and Speaker Diarization

Apache License 2.0

1.02k stars 89 forks source link

sv-rdino trainning data. #80

Closed hcfeng201 closed 5 months ago

hcfeng201 commented 5 months ago

Regarding sv-rdino, how many data(persons and utts) are needed to achieve an EER below 10%? Have you conducted any experiments on this?

yfchenlucky commented 5 months ago

According to VoxCeleb, the Equal Error Rate (EER) typically falls below the 10% threshold after approximately 10 training epochs. For other datasets, no relevant conclusions can be drawn regarding this trend.