fnzhan / UNITE

[CVPR 2022 Oral] Marginal Correspondence for Conditional Image Generation, [CVPR 2021] Unbalanced Feature Transport for Exemplar-based Image Translation
192 stars 26 forks source link

Question with lpips results in your MCL-net paper #11

Closed mlyarthur closed 1 year ago

mlyarthur commented 2 years ago

image I have a question about your lpips score calculation. Should'nt lpips score is the lower the better? Does you adopt different metric calculation?

fnzhan commented 1 year ago

Here LPIPS is used to evaluate the diversity (not quality) of different exemplars.