search

nihaomiao / CVPR23_LFDM

The pytorch implementation of our CVPR 2023 paper "Conditional Image-to-Video Generation with Latent Flow Diffusion Models"
BSD 2-Clause "Simplified" License
443 stars 42 forks source link

cFVD and sFVD? #3

Closed jacob-kang closed 1 year ago

jacob-kang commented 1 year ago

Thanks for your great work.

I saw your paper, and I want to enquire what is difference between cFVD,sFVD and normal FVD. And where can I find cFVD, sFVD metric code?

Thanks.

nihaomiao commented 1 year ago

Hi, @jacob-kang, cFVD and sFVD can be easily computed based on normal FVD. For example, if you have 10 classes, you can first generate 10,000 videos for each class and then compute FVD for each class. The final cFVD will be the mean and std of these 10 FVD scores. The computation of sFVD is also similar. Just compute FVD for each subject and report the mean and std as the final sFVD score.

jacob-kang commented 1 year ago

Oh I got it. Thanks. I close this issue.