An algorithm that mimics the styles of some of history's greatest painters has been developed by researchers in Germany.
In the study, which has been submitted to Nature Communications, a complicated mathematical code was devised in order to create a "convolutional neural network", capable of turning images into imitation works of art in just an hour.
By taking an existing photograph -- of a person or building -- and merging it with a painting, such as a Picasso or Van Gogh, the network uses object recognition to recreate the image in line with the features of the artwork. The result? You can have your home immortalised in the hypnotically bleak style of Munch's The Scream, or create a Cubist self-portrait.
The technique is impressive in itself but it's not simply another case of machines showing how they can out-dazzle humans at their own game. Rather, it's designed to demonstrate exactly how these types of neural networks can separate an image's style from its content.
The paper's authors, led by University of Tuebingen PhD student Leon Gatys, explained: "The key finding of this paper is that the representations of content and style in the convolutional neural network are separable. That is, we can manipulate both representations independently to produce new, perceptually meaningful images."
It's a delicate balancing act to ensure the recreations are successful: too much focus on content and the image looks too bland; too much emphasis on style and the output image is too abstract.
The network can only copy existing styles, rather than creating its own original works, meaning that it may be some time before we see a computerised Picasso hanging in the Tate.
This article was originally published by WIRED UK
