search

wswdx / C2F-FWN

data/code repository of "C2F-FWN: Coarse-to-Fine Flow Warping Network for Spatial-Temporal Consistent Motion Transfer"
MIT License
51 stars 8 forks source link

Quantitative metrics #6

Open FrankMelon1 opened 3 years ago

FrankMelon1 commented 3 years ago

Thanks for your great work!

Could you share the SSIM, PSNR, LPIPS, FID, TCM metrics code?

I am looking forward to your reply, thanks!

wswdx commented 3 years ago

Our evaluation code is not clean yet and relied on many third-party packages. We may release them after after clean-up and with documents in the future. However, it's easy to follow existing works to do the evaluation. SSIM and PSNR are evaluated using scikit-learn, LPIPS is evaluated based on this repository (https://github.com/richzhang/PerceptualSimilarity), FID is evaluated based on this repository (https://github.com/mseitzer/pytorch-fid), TCM is based on our own implementation with the help of FlowNet-v2 (https://github.com/NVIDIA/flownet2-pytorch).