New algorithm gives photos Picasso-style makeovers
Source: Adario Strange

Google has already shown us the weird possibilities when images are processed through a neural network.

But a new experiment takes this dynamic to a new level by transforming any photo into an image that mimics the painting style of one of the old masters.

See also: Google Maps code transformed into infinite series of abstract art

The details of the project are revealed in a research paper titled "A Neural Algorithm of Artistic Style."

"Here we introduce an artificial system based on a Deep Neural Network that creates artistic images of high perceptual quality," reads the paper, penned by a group of researchers from the University of Tubingen in Germany.

"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."



Image: Gatys, Ecker, Bethge

Using an image of a street in Germany (above), the team demonstrated the ability to recreate the photo using the visual styles of Vincent van Gogh, Edvard Munch and Pablo Picasso. Each transformed street photo, at least upon casual inspection, looks like it was indeed painted by one of the masters.



Image: Twitter screenshot via Andrej Karpathy

Over the last couple of days, Andrej Karpathy, a PhD student at Stanford studying Machine Learning, posted a few experiments on Twitter using the technique, and the results were incredibly accurate. Karpathy had since removed all examples of his experiments with the image transfer, but we managed to capture some screenshots of his tests, including one that turned a photo of Gandalf into a Picasso-style portrait.



Image: Twitter screenshot via Andrej Karpathy

Does this mean that, with a little more code tweaking, we'll begin to see fake works (created using this technique) suddenly "discovered" in coming years? Probably not. (Remember, there's still the issue of paint and canvas to consider.)



Image: Twitter screenshot via Andrej Karpathy

But it does mean that machines have just inched a little closer to putting artistic imitators out of work, as well as giving us a window into how Deep Neural Networks (in this case, Convolutional Neural Networks) may transform how we see the world in the near future.