Artificial intelligence detects false works with a brushstroke
The strokes have to be clearly delimited
On November 15, an art collector paid nearly 400 million euros for a Da Vinci 65x45 cm. An investment that adds to the billions of euros that moves the Art market in everything in the world.
Hundreds of transactions with strict security controls to avoid sales of fraudulent works. Identifying these fakes is not simple or cheap. The main techniques include infrared rays among other tools.
However, American and Dutch researchers have managed to develop an artificial intelligence device capable of detecting the authenticity of the works just by their strokes.
To learn how to detect fake works, the University of Rutgers (United States) and the Study of Restoration and Research of Paintings (The Netherlands) have built a database with more than 300 drawings of famous artists.
To carry out this project, the researchers created a neural network capable of detecting the specific characteristics of each artist to identify their strokes. Along with this, they developed an algorithm to look for these characteristics.
After the checks, the project leaders found two different systems to detect fake works, but the combination of both offers a very powerful detection method
With both algorithms working in tandem, the researchers managed to correctly identify the artists about 80% of the time. "A human can not do that," says the professor at Rutgers University and one of the authors of the article, Ahmed Elgammal.
This technique, however, is only effective when the lines are clear and defined, so it is useless for works whose lines are blurred.