Artificial Intelligence to detect the fakes
The detection of fakes in art has always posed a thorny problem for many specialists who have often used their flair to form opinions, sometimes wrong, unfortunately. Still, artificial intelligence could come to their rescue and end disturbing controversies with an application developed by researchers at the Rutgers University of New Jersey and the Dutch Painting Restoration and Research Service. These published the results of their experiments in a dossier titled "Picasso, Matisse or a fake ?, Automatic Analysis of Drawings at the Brushstroke for Attribution and Authentication" through an application called AI that compares to the extreme of the keys used to compose an image. In short, an artist's keys would be like his fingerprints, taken from a line deriving from the movement of his hand, which a specialist is not necessarily able to detect at a glance. In contrast, others call on laboratories to carry out a battery of scientific tests on a canvas that can be conclusive concerning its dating and the pigments used without it being possible to determine conclusively that the key is indeed the hand of its author. With AI there is practically no more possible error. Drawings will be understood by AI, as we have seen with the examination of works by Picasso, Matisse or Egon Schiele since they are more than 80,000 keys subjected to an analysis carried out with algorithms that detected notable differences, such as in the weight of a trace. In most cases, the machine can correctly identify a work and rule out plagiarism just by focusing on one touch, which a human being cannot do. For works on paper, AI also has the advantage of being an inexpensive method compared to analyzes carried out in the laboratory. In contrast, if a forger can sometimes draw inspiration from a great artist, it will be impossible for him to replicate his hand weight on the keys that ultimately represent his autograph. Nevertheless, Al has its limits since we stayed with the drawings. Researchers are currently trying to test this application for Impressionist and modern painters' oils through analyses of brushstrokes to go further than the laboratory methods.
© Adrian Darmon
Artificial Intelligence explained, video
The detection of fakes in art has always posed a thorny problem for many specialists who have often used their flair to form opinions,...