Dostoevsky in color: a portrait of the writer from 1879 was restored using a neural network

Stable Diffusion appeared in the neural network subreddit publication, where user FotoRe_store showed and described the restoration of a portrait of Fyodor Mikhailovich Dostoevsky. The photograph was taken as a basis from 1879.

The author noted that he tried to make the image as close as possible to the original, that is, it was a restoration, not a reconstruction. At the same time, special emphasis was placed on the texture of the facial skin so that the result was as realistic as possible.

To do this, the author systematized a lot of data about the author’s appearance during the period when the original photograph was taken. He also studied earlier colorizations of this photograph and other portraits of Dostoevsky.

He used the Stable Diffusion neural network to increase the resolution of the digital version of the image and many sketches of individual fragments of the portrait. The author compared this whole process to assembling a puzzle from a thousand pieces of the same shade based on a vague picture of the original, which exists only in the mind of the person putting it together.

How do you like the result? Write in the comments.