About the LPIPS evaluation metrics in Tab3

[csguoh]

Hi, authors!

Sorry to bother you again. In tab3, I noticed that you have used the LPIPS as one of the evaluation metrics. However, to my knowledge, the LPIPS is the LOWER the BETTER. If it is true, the reported results actually shows that the proposed HRA is inferior to other methods since it possess a lager LPIPS. Is something wrong with my understanding?

Thx

[DaShenZi721]

Thanks for your appreciation of our work! Actually, we use the perceptual metric (LPIPS) to evaluate sample diversity. That is to say, a higher LPIPS score indicates greater diversity in the generated images.