Hi, thanks for open-sourcing this project.
I am trying to reproduce the evaluation results in the paper. Currently, I can reproduce PSNR, SSIM, LPIPS, and IDS, but I am unable to reproduce AKD, FVD, and TLME.
I have a few questions regarding the implementation details:
AKD: How exactly is AKD computed in your paper? Is there any reference implementation or official code for it, including the landmark detector, normalization, and averaging strategy?
FVD: Which FVD implementation did you use (for example, the StyleGAN version)? Is FVD computed at the dataset level, or per video and then averaged?
TLME: How is TLME defined and computed? Is there any released code for this metric?
I saw that the paper mentions metric details in the supplementary material, but I could not find the supplementary material or any public implementation details for AKD, FVD, or TLME in the released code.
Any clarification would be greatly appreciated. Thanks!
Hi, thanks for open-sourcing this project.
I am trying to reproduce the evaluation results in the paper. Currently, I can reproduce PSNR, SSIM, LPIPS, and IDS, but I am unable to reproduce AKD, FVD, and TLME.
I have a few questions regarding the implementation details:
AKD: How exactly is AKD computed in your paper? Is there any reference implementation or official code for it, including the landmark detector, normalization, and averaging strategy?
FVD: Which FVD implementation did you use (for example, the StyleGAN version)? Is FVD computed at the dataset level, or per video and then averaged?
TLME: How is TLME defined and computed? Is there any released code for this metric?
I saw that the paper mentions metric details in the supplementary material, but I could not find the supplementary material or any public implementation details for AKD, FVD, or TLME in the released code.
Any clarification would be greatly appreciated. Thanks!