Neural network taught to restore the “pixel” photos

Russian developer Denis Malimonov created an application based on AI algorithms that can convert very low resolution “pixel” images into realistic portraits. Enthusiasts have already tested the development called Face Depixelizer in action – as it turned out, artificial intelligence is able to work not only with portrait photos, but also with images of the heroes of classic video games.

When creating the program, the possibilities of StyleGAN were used – a tool that, with the help of machine learning, can generate realistic photos of non-existent people. Having received the “pixel” source, the application analyzes the picture and improves its clarity, completing the missing details.

Processed persons do not have to belong to real people – the algorithm can be used to see how the characters of old video games will look in real life.

However, it is worth noting that the current version of Face Depixelizer is still far from ideal: the results of depixelization sometimes turn out to be not only very different from the original, but also slightly frightening.

The developer has already published an online version of the application in the public domain – anyone can get acquainted with its capabilities.

Leave a Reply

Your email address will not be published. Required fields are marked *

12 + seventeen =