search

Doubiiu / CodeTalker

[CVPR 2023] CodeTalker: Speech-Driven 3D Facial Animation with Discrete Motion Prior
MIT License
515 stars 57 forks source link

my result is not the same as the paper result. why? #36

Closed zhengshi119 closed 1 year ago

zhengshi119 commented 1 year ago

Frame Number: 3879 Lip Vertex Error: 1.8149e-03 FDD: 4.6162e-05

Doubiiu commented 1 year ago

Hi. I didn't come across that high LVE value in the experiments. If you train the model from scratch, the differences may come from: (i): the preprocessed dataset (you can test by using the provided model) (ii): You can test the intermediate models which may have a better balance between perceptual quality and quantitative performance. The reasons are various, I am not sure about that.

zhengshi119 commented 1 year ago

thank you！