A painting by Renoir authenticated thanks to artificial intelligence | Knowledge of the Arts

Faced with the uncertainty of specialists, a Swiss collector had a painting authenticated using artificial intelligence. According to the algorithm, the painting would have an 80.58% chance of being by the hand of Pierre-Auguste Renoir.

On November 19, the British media ” The Guardian revealed that an array of Pierre-Auguste Renoir (1841-1919) would have been authenticated by an artificial intelligence. Faced with the indecision of specialists in Wildenstein-Plattner Institute, an art history research center, the owner of the painting turned to the Zurich start-up Art Recognition. This company specializing in the use of artificial intelligence applied to art has analyzed Portrait of a woman (Gabrielle) and concluded that there would be an 80.58% chance that the painting was by the hand of the Impressionist painter. Experts from Dauberville & Archives Bernheim-Jeune, which publishes a catalog raisonné of known works by the artist, concluded that it was indeed a Renoir after analyzing the pigments in the painting.


Artificial intelligence, a future tool for experts?

How could artificial intelligence determine that this half-length portrait of a young woman with dark hair dressed in a white shirt whose features resemble Gabrielle Renard, the nurse of Renoir’s son and one of his models favourites, was indeed from the hand of the Impressionist master? By comparing high-definition photographs of the painting and its details to those of 206 authenticated paintings by Renoir, the algorithm studied the style, the shape of the brushstrokes and the combinations of colors complementary to each work.

Portrait of a Woman (Gabrielle) was analyzed by an artificial intelligence developed by Art Recognition before being attributed to Pierre-Auguste Renoir. © Art Recognition

However, this technology has its limitations. This clever calculation does not yet take into account the possible poor condition of the canvas, whether due to dirt, overpainting or more or less successful restorations, which can then distort the data. Likewise, the results of the artificial intelligence do not take into account the scientific analysis of the support and the pigments used by the artist. Also, the precision of the algorithm is dependent on the quality of the paintings on which it was trained. If a fake or a copy is part of the corpus or if one of the selected works is damaged, the result is misleading and the degree of certainty is much lower.

Pierre-Auguste Renoir, Gabrielle with a rose, 1911, oil on canvas, Musée d'Orsay.  ©Wikimedia Commons/The York Project

Pierre-Auguste Renoir, Gabrielle with a rose, 1911, oil on canvas, Musée d’Orsay. ©Wikimedia Commons/The York Project

More than 500 authentications against this artificial intelligence

Until today, the Swiss company has already carried out more than 500 authentications with this artificial intelligence and trained its algorithm to recognize 300 artists, whether they are impressionist painters or old masters. This fallArt Recognition announced through its tool that the only Titian (1488-1576) kept in an art museum in Switzerland (in Zurich Kunsthaus), Evening landscape with couple of figures (1518-1520), was probably not painted by the Italian artist of the Venetian school.

If this new technology does not claim to completely replace the experts, it can nevertheless help them to authenticate or not a work: “ Artwork owners are often told by connoisseurs that it is their “feeling” or “gut feeling” that determines whether a painting is authentic or not, which can be very frustrating.explains Dr. Carina Popovici, CEO of Art Recognition, during a meeting on the use of the technology in the art trade at the Art Loss Register in London. They really appreciate the fact that we are more specific “.

We want to thank the author of this write-up for this outstanding content

A painting by Renoir authenticated thanks to artificial intelligence | Knowledge of the Arts


Visit our social media profiles and other pages related to themhttps://www.ai-magazine.com/related-pages/