search

BadToBest / EchoMimic

Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning
https://badtobest.github.io/echomimic.html
Apache License 2.0
2.26k stars 263 forks source link

About FID/FVD metric #56

Closed Vincent-luo closed 1 month ago

Vincent-luo commented 1 month ago

Thanks for your great work! I have a question about the model evaluation. Many related works report FID/FVD metrics for their models, but they often lack specific details about the evaluation process. I'd like to know whether you calculated FID/FVD on the training split or the test split, and how many images/videos you generated when computing these metrics. If you could provide more specific evaluation details, it would be helpful for fair comparisons between models.

JoeFannie commented 1 month ago

We will consider adding the details to our paper. Thanks for your advise.