search

jychoi118 / P2-weighting

CVPR 2022
MIT License
143 stars 14 forks source link

FID score for FFHQ #6

Closed napCC closed 1 year ago

napCC commented 1 year ago

I tried to reimplement the FID score for FFHQ by using code from torch-fidelity. However, the score is higher than the paper mentioned. I notice the evaluations folder does not have the FFHQ reference batch. Can you please provide the fid score code and the real image used in the evaluation?

jychoi118 commented 1 year ago

We measured FID scores with clean-fid First install the library: pip install clean-fid Then, you can measure FID for FFHQ by running: from cleanfid import fid score = fid.compute_fid(fdir, dataset_name="FFHQ", dataset_res=256, dataset_split="trainval70k") You should put the directory of your samples in fdir.

napCC commented 1 year ago

Thanks