Deep learning algorithms have found a remarkable number of applications within just a few brief years: Google has used it for its image recognition and translation tools, and medical companies are developing programs that can detect diseases.
Now, it’s being applied to the replication and understanding of art. A group of researchers at the University of Tubingen in Germany developed an algorithm that can distill the essence of a painting’s “style” and transform any image to fit that style, allowing the computer program to “paint” after the fashion of an inputted artwork.
“The system uses neural representations to separate and recombine content and style of arbitrary images, providing a neural algorithm for the creation of artistic images,” the researchers wrote.
A photograph of a row of apartments by the river in Tubingen, Germany were filtered separately through the style of a series of a paintings, including J.M. Turner’s “The Wreck of a Transport Ship,” Van Gogh’s “The Starry Night,” and Edvard Munch’s “The Scream.”
The key achievement of the algorithm was that neural networks could learn to separate the content and style of an image. In the transformed images, the arrangement of the objects in the original photograph were largely preserved, with the “colors and local structures” morphed to emulate the inputted artwork.
However, the delineation between style and content is incomplete. The more faithful the emulation of an artwork’s style, the more content of the original photograph was distorted or lost, and vice versa, and the right balance had to be found on a sliding scale by hand.