In order to give you a better service Deutsche Messe uses cookies. If you continue we assume that you consent to receive cookies on all Deutsche Messe websites. Legal Notes

Artificial Intelligence

Drawing the line at art forgery!

A study by researchers from Rutgers University in New Jersey and the Atelier for Restoration & Research of Paintings in the Netherlands shows how AI can analyze the individual brush or pencil strokes in a work of art so accurately that the artist behind it can be identified.

13 Dec. 2017
Rutgers AI Art

It was an incredible scene - after a bidding war lasting not even twenty minutes at Christie's in New York, the "Salvator Mundi" painting, believed to be by Leonardo da Vinci, was auctioned for more than $450 million. Even though that includes all the fees, it can hardly be a called a meager sum. But could it later prove to have been a bargain? Nobody can foresee the way the overheated art market is going to develop next - even artificial intelligence (AI) systems can only make predictions. However, when it comes to identifying forgeries in art, AI could soon deliver an effective alternative to conventional methods used by art experts such as infrared spectroscopy, gas chromatography and radiometric dating.

Researchers from the Department of Computer Science at Rutgers University in New Jersey and the Atelier for Restoration & Research of Paintings in the Netherlands have developed an AI system that can already identify 80,000 individual strokes by famous artists including Picasso, Matisse and Modigliani. This is thanks to the system's analysis of 300 line drawings, with its deep recurrent neural network (RNN) helping it to learn which features in the strokes are specific to which artists.

To be on the safe side, the researchers also trained a machine learning (ML) algorithm to look for specific features in an artwork, like the shape of the line in a stroke. By combining these two methods of analysis, artists can be identified correctly in some 80 percent of cases. Even more impressively, the RNN can clearly detect characteristics in an original that a forger has not noticed, as well as characteristics in a forgery that are not present in verified originals by a particular artist.

Already a major cause of concern for the world of art forgery, the AI system can currently only be used for drawings in which the lines are distinct. However, the researchers are already planning to test their results further using Impressionist paintings and other 19th century works featuring clearly identifiable individual brush strokes. All in all, CeBIT 2018 with its focus on AI is set to be an interesting event for art experts and historians as well as the business community.


Department of Computer Science, Rutgers, The State University of New Jersey (Piscataway, NJ 08854-8019, USA)


Related Articles