https://phillipi.github.io/pix2pix/
image to image translation / guided art synthesis
If i get it right, that sounds like a ridiculous tool! I can’t believe you can input one of the “input” images and output one of the “output” images. If it does work, this thing is clearly wizardry of the illuminati.
I think that they have a bank of images with their parts labeled
‘doorframe’ window windowframe etc, and when you feed it colors, they represent an area you are trying to fill with one lf those things,
its like a selective style transfer with a source material set,
I wonder if it can spit out any color of the same building?
Wow! If it even works little bit like on the Screenshot, it is pretty amazing! Thanks for sharing
My bullsh1t detector went full scale for a moment…
From the 5 minutes I browsed the host domain, references and code:
this seems to “legit” but advertised in way to make it seem to do more things than it actually does.
It feels like a case of no-mans-sky.
or maybe I also failed the IQ test here?