[eagsalazar2]

When I saw "restoring" in the title I was expecting higher resolution. For example seeing in modern photo level detail eyelashes, wrinkles, etc. I get that, like the colors, this would require the adding lots of made up information about scene and feature details but IMO it would blur the lines between restoration and reconstruction/storytelling in a really awesome way. Old photos are cool in their own way but their lack of detail makes them seem so alien. Would be exciting to get a hyper real reconstruction.

Are there examples of ML doing something like that? (also know little about ML)

[citnaj]

Author here: It's early, and currently resolution is limited primarily by model size. Which drives me nuts. It's one of my top priorities to address because that would be a great improvement. Adding super-resolution to the pipeline should also be pretty easy but I want to at least output a reasonable base resolution on the photos first before I go that route.

Oh yeah to answer your question- super resolution does indeed make up details as you describe there and arguably does blur the line with restoration/story telling. But so does colorization- not all the colors added by the model are going to be what was actually going on there, of course.

[slg]

There reminds me of a recent 99% Invisible episode [1] in which they discuss the same topic in the world of dinosaurs. It details how dinosaurs used to be depicted with the goal of only showing the things that we are confident in being true (although what we are confident in obviously changes over time). This results in mostly just greenish-brown skin draped over a muscle structure attached to the fossilized skeletons.

In recent decades there has been a push to show the animals more realistically. The fossilized evidence is studied and compared to the skeletal structure of animals that exist today. Inferences and educated guesses are made from there to project a more realistic but more subjective image of the dinosaurs. We now get much more varied and interesting depictions with feathers, bright coloring, fat deposits, and other features that can neither be completely confirmed or ruled out based on the evidence.

[1] - https://99percentinvisible.org/episode/welcome-to-jurassic-a...

[eagsalazar2]

Hah yeah I just listened to that a couple days ago but didn't make the connection. Probably was rattling around my subconscious when I wrote this question because yeah it is very similar. That's a particularly interesting comparison too because the whole point was that just filling in conservatively based on experience misses a ton a real-world crazy and interesting diversity. The best example was how if we were imagining what elephants looked like just based on their fossilized skeletons, they wouldn't have trunks!

[gowld]

https://www.youtube.com/watch?v=gg0F5JjKmhA