search

YuanxunLu / LiveSpeechPortraits

Live Speech Portraits: Real-Time Photorealistic Talking-Head Animation (SIGGRAPH Asia 2021)
MIT License
1.2k stars 209 forks source link

GMMLogLoss for training audio2headpose #92

Open shanzhajuan opened 1 year ago

shanzhajuan commented 1 year ago

Nice work. I have two questions while training my own audio2head network:

  1. I noticed that the GMMLogLoss is negative. Is this correct?
  2. Does the training data need to be preprocessed, like normalized, adjusted mean, etc.. In my training data, the angles are in degree format. Rotations and Translations are not normalized. Is this correct?