Picking the best colour is a difficult problem, I don't think there's a good general answer short of ML magic. I assembled an adversarial collection of images:
For every heuristic, I can think of an image that breaks it. On the other hand, I just wanted to do better than the 1x1 trick, and I think the library clears that bar.
alt Hacker News
Picking the best colour is a difficult problem, I don't think there's a good general answer short of ML magic. I assembled an adversarial collection of images:
https://github.com/si14/okmain/blob/main/test_images/IMG_134...
https://github.com/si14/okmain/blob/main/test_images/pendant...
https://github.com/si14/okmain/blob/main/test_images/pendant...
https://github.com/si14/okmain/blob/main/test_images/red_moo...
https://github.com/si14/okmain/blob/main/test_images/supremu...
For every heuristic, I can think of an image that breaks it. On the other hand, I just wanted to do better than the 1x1 trick, and I think the library clears that bar.