caught in motion, and it caught me too.
Your photo goes through CLIP, an image model, and then a small model trained on about 3,000 of my own A/B choices between 500 photos. It predicts my taste. In testing it agreed with my choice about 2 times in 3 (Spearman ρ = 0.45).
This guesses whether I, Hamadullah, would like your photo. It's trained on my own taste, so it says nothing about whether your photo is objectively good.
In testing it agreed with my choice about 2 times in 3: better than chance, far from perfect. Take the score as a fun opinion, not a verdict.
Your photo is scored and then forgotten. Nothing is saved.