This System Converts Sketches into Photorealistic Portraits

 - Jun 17, 2016
References: engadget & petapixel
Designed by Yagmur Gucluturk and Umut Guclu at Radboud, this neural network can interpret and convert sketches into computer-generated realistic portraits.

The technology was created by having thousands of images inserted into the computer along with grayscale sketches of them in order to teach the computer how to convert those sketches into realistic looking portraits. The end result of the experiment turned out to do just that, with the network's computer-generated portraits looking almost as real as the photos they were inspired by.

This neural network has several implications, one of which is that it proves and shows the potential of the idea that computers can learn information and apply it autonomously. As a system in itself, these computer-generated portraits can have applications in the security world by producing photorealistic images of criminal sketches.