"To better understand your FID and KID calculations, could you specify the data you used as the reference set for these metrics? I see that the generated data from your algorithm was included, but I’d like to know more about the nature and characteristics of the other data used for comparison — for example, was it the20,000 high-quality, high resolution(HR) face images from the real-world FFHQ
dataset, or another curated dataset? Additionally, did you perform any preprocessing on this reference data to align it with the generated data for a more accurate metric evaluation?"
also the data to test NIQE metric for your generated data against it
Thank you in advance...