There appears to be a new painting by Italian Renaissance great Raphael, and we can thank an unusual source: facial recognition technology.
Based on similarity of previously unsigned de Brécy Tondo and Raphael’s Madonna face sistine madonna, a team of researchers from the Universities of Nottingham and Bradford in the United Kingdom has determined that there is a 97 percent probability that the enigmatic circular painting, which has been studied by academics and art historians for decades, is the work of the famous Old Old. Master, BBC reports.
The researchers compared the two works using an artificially intelligent facial recognition system that Hassan Ugail, a professor of visual computing at the University of Bradford, began developing in 2002.
A deep neural network, trained by machine learning to process data in a hierarchical structure, performs a direct facial comparison of two paintings and finds a statistically high probability that the same artist created both works. The two Madonnas are found to be 97% similar, and the baby Jesus is 86% similar. A similarity of 75 percent or more is considered identical, the researchers claim.
“Human eyes see obvious similarities in faces, but computers can see much deeper than we can, down to the pixel level in thousands of dimensions,” Ugel said in a statement. “My co-authors and I concluded that The conclusion is that the same model was used for both paintings, and there is no doubt that they are by the same artist.”
The use of facial recognition and other forms of machine learning can prove to be a powerful tool for artwork appraisal.
“Technology can reveal what’s hidden,” Adam Lowe, director of art and digital technology studio Factum Arte and founder of the Factum Foundation, told Artnet News in an email. “I doubt that studying surfaces will prove more accurate than facial recognition. A painting is a representation, in this case of a face – a three-dimensional form by painting with a brush. There are always traces of style, and Pattern recognition software can help reveal these.”
Lowe has been using machine learning to identify signature marks in the way artists paint on surfaces. “This used to be called the artist’s style, but really, it was information embedded in the surface of the painting,” he said.
An AI analysis of de Brécy Tondo further supports the late British art collector George Lester Winward’s long-held belief that the painting, which he bought at a country house auction in England in 1981, is lost Raphael masterpiece.
The Old Masters Gallery of the State Kunstmuseum in Dresden, featuring sistine madonna– is actually best known for its two little angelor Baby Angel, at the bottom of the canvas – it has been previously determined that the likeness to de Brécy Tondo is a late copy of the original, possibly from the Victorian era.
The new findings complement the conclusions of other studies of the painting conducted through Winward’s De Bracy Trust, including scanning electron microscopy and Raman spectroscopy, which found its material composition to be consistent with Renaissance paintings.
The research, published in the academic paper “Deep Facial Features for Analyzing Artistic Depictions,” was announced in a presentation at the Software, Knowledge, Information Management and Applications Conference at the Cambodia University of Science and Technology in Phnom Penh in December.
Follow Artnet News on Facebook:
Want to stay on top of the art world? Subscribe to our newsletter for breaking news, eye-opening interviews and sharp criticism to move the conversation forward.