Hacker News new | past | comments | ask | show | jobs | submit login

Idea for computer-imagery/ML people: a method to realistically recolor old photos by sampling from new photos of the same place, with some caveats (for eg. the object must have a corresponding object, etc)

I remember seeing this/such papers elsewhere. They have appeared in conference/papers for the last few years too. Here I'm suggesting untrained coloring of 'new' images from existing though, so maybe what I'm asking is a simpler problem? "Transferring colors from base (new) image to target (old) image using such and such sampling" would be how I would put my problem. The paper you link to would be more "learning to transfer colors by looking at how grayscale images transfer to color", right? Am I missing something?

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact