Re:Sharing learnings about our image cropping algorithm
by fuzzythinker on Wednesday May 19, 2021 @05:59PM
(#27215078)
Breasts shouldn't have more entropy than face. Perhaps the reason is due to the breasts being in the middle of the picture, so the face gets being compared to bottom rows more frequently?
[ Parent
| on HN ]
Re:Sharing learnings about our image cropping algorithm
by ad404b8a372f2b9 on Wednesday May 19, 2021 @07:18PM
(#27215666)
Why not? Shirts might have flashy patterns, differently colored fabrics, alternating skin and shirt. On a row-per-row basis I can see the chest area being more entropic than a face with an even skin tone.
edit: I googled "woman" and selected random pictures which showed the whole upper body, entropy summed over each row to the right: https://imgur.com/a/oVB57gu
Re:Sharing learnings about our image cropping algorithm
by LorenPechtel on Wednesday May 19, 2021 @10:31PM
(#27216695)
Seconded. I'm picturing my wife's choices in clothing--I would say often the highest entropy will be her top. And depending on the resolution and how good the entropy measurement is I can easily see her pants coming in second. (She tends to favor repeating patterns.)
(She has no work dress code she needs to conform to. Rarely does she wear a top as plain as any of your examples.)