Evaluation score between my retraining and CodeTalker paper is not the same.

Doubiiu / CodeTalker

[CVPR 2023] CodeTalker: Speech-Driven 3D Facial Animation with Discrete Motion Prior

MIT License

531 stars 58 forks source link

Open youngstu opened 1 year ago

youngstu commented 1 year ago

Evaluation score between my retraining and CodeTalker paper is not the same.

The first stage model was reused and the second stage model was retrained, but the final score were inconsistent with those in the paper.

Could you help explain? Thanks.

Retraing result:

Frame Number: 3879
Lip Vertex Error: 5.2776e-04
FDD: 4.4944e-05

Paper result: Screenshot from 2023-04-25 16-35-24

ffxzh commented 1 year ago

yep, I have the same question Frame Number: 3879 Lip Vertex Error: 6.0386e-04 FDD: 4.3230e-05

FarzanehJafari1987 commented 3 months ago

How do you preprocess the BIWI dataset? Your results are very close to those reported in the paper, but mine differ significantly.