Scientists at the Radboud University have been working on a model that uses an algorithm that can turn drawings and sketches of faces, and even pieces of art, into photorealistic images.
The new machine-vision algorithm relies on neural networks to turn hand-drawn sketches of faces into photorealistic portraits.
In order to teach the algorithm to turn a sketch into a photo, the researchers used a dataset that had sketches and photos that matched (so a photo and its sketch). The algorithm then received only the sketches and started turning them into images. Obviously, at first, the algorithm couldn’t make images identical to those in the dataset, but in time, it started to learn on its own.
Essentially, this machine inverts a face sketch to synthesize photorealistic face image, thanks to its deep neural networks.
The new smart algorithm will be used in forensics, to turn sketches based on eyewitness accounts into realistic photos.
Interestingly, this algorithm can also be used in fine arts, as it can turn self-portraits into real photographs.