DeOldify A Deep Learning based project for colorizing and restoring old images (and video!)

New and Exciting Stuff in DeOldify
Glitches and artifacts are almost entirely eliminated
Better skin (less zombies)
More highly detailed and photorealistic renders
Much less "blue bias"
Video - it actually looks good!
NoGAN - a new and weird but highly effective way to do GAN training for image to image.

DeOldify.NET executable file for Windows and Linux(Default exe was use Artistic.hmodel)
or you can compile youself for better result

some test

use Artistic.hmodel

use Stable.model

Would be cool when David implements AI into G'MIC to do such things; time will tell.

Thank You for providing information about the program.
Unfortunately, my computer does not meet the hardware requirements - a minimum of 4 GB of memory for the graphics card (with my 1 GB card).
Installed the latest version 2.1 shuts down after about ~90% of the task.
So I downloaded the older version 1.0 (it's a shame it only uses the artistic model though) and it works for my slow machine.
https://github.com/ColorfulSoft/DeOldif ... s/tag/v1.0
For Windows, it's not an installer, but a 123 Mb executable (it starts slowly for me ~25 s).


I tried the same examples as posted and got this result:







For the test, I compared the original photos converted to B&W with the result from DeOldify.
A bit surprising at times, but very interesting to me.

Original image:


Black & White G'mic - default settings:


Result from DeOldify V-1.0: