An algorithm developed at a German University is able to imitate the artistic styles of some of the world’s most famous painters.

In a study, titled A Neural Algorithm of Artistic Style and published in the journal Nature Communications, the research team describes a complex mathematical code which generates a “convolutional neural network” to produce “high perceptual quality” paintings, mimicking great artists’ styles, in less than an hour.

The paper features example images which borrow the style of Vincent Van Gogh, Pablo Picasso, self-portrait artist Frida Kahlo, and The Scream painter Edvard Munch. One example shows how Van Gogh’s 1889 The Starry Night, is merged with a night-time shot of Stanford University campus:


San Francisco’s Golden Gate bridge formed the subject matter in a further test – its results are shown below:


The artificial system takes a simple photograph of an object, landscape or person, aligns it with a masterpiece, and uses recognition technology to redesign the image mirroring famous techniques. According to the paper, the algorithm allows a user to ‘trade-off’ the weight of the style and to modify levels of content reconstruction as desired.

Leon Gatys, lead author and University of Tuebingen PhD student, explains: ‘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.’

He adds: ‘…in light of the striking similarities between performance-optimised artificial neural networks and biological vision, our work offers a path forward to an algorithmic understanding of how humans create and perceive artistic imagery.’

This is yet another example of how computers are encroaching on supposed unconquerable human capabilities. For now though, the network can only imitate the designs and styles of others with no plans to attempt an original any time soon.