search

deepaudio / deepaudio-speaker

neural network based speaker embedder
25 stars 5 forks source link

Pretrained models? #23

Open hbredin opened 2 years ago

hbredin commented 2 years ago

Salut Ruiqing ;-)

I was wondering whether

yinruiqing commented 2 years ago

Salut Hervé. Long time no see ;-)

  • you had plans to share weights of models trained with this project Yes. But I don't have my own GPU now. I will share the weights when I can train the model locally (maybe next month with a 4090) :)
  • you could report performance on VoxCeleb I trained ECAPA several months ago. The eer is around 1.13% on vox-o. It should get better performance on clean version. I will report the result when I finish the training locally.
zuowanbushiwo commented 1 year ago

@yinruiqing Any updates? Thanks！