Could an agent no develop a bespoke correction algorithm for the photos of the photo?
Why feed it photos of a photo when you could just feed it the photo that was printed out?
alt Hacker News
Why feed it photos of a photo when you could just feed it the photo that was printed out?