The meaning of the output score

#2
by Rurrres - opened

Hello @xswu , thank you very much for this demo. I read the paper but I can not understand the meaning of the output HPS v2 very well, in the paper it was explained as "HPS v2 estimates the probability of an image to be preferred by viewers against other images", but the output seems always to be 0.2xx, does this mean that there is a chance of 0.2xxx that this image wins the pairwise comparison?

The number is not a probability, but a score. The model is trained to give a higher score for preferred images, but the absolute value does not have a physical meaning. I will revise the paper about the statement. Thank you for your question!

Sign up or log in to comment