For more than 40 years researchers and scholars of art history have looked at the Bresi Tondo, a 95 cm diameter circular painting belonging to the Bresi Trust, with more than admiration, emotion or pure pictorial enjoyment. He has looked at it with curiosity. An urgent and concrete curiosity which is proved in a relatively simple question: who is its author? Was it painted by Raphael, as collector George Lester Winward stubbornly believed when he bought the piece in the early 1980s?
Simple question to ask, of course. Answering it is another story. For decades this question is being thrown around and no one is able to give a concrete answer. So far. Thanks to the support of artificial intelligence, a group of researchers claim to have discovered once and for all a strong answer,
To understand the result it is necessary to know the painting in advance. And the reason is Rafael’s theory. In December 1981 George Lester Winward, a businessman from the British county of Cheshire, decided to purchase the piece and add it to his ambitious art collection. He did not do so because of his crush on the painting and its representation of the Virgin Mary with baby Jesus in her arms. Or not just for him, at least.
see further than the eye can see
A detail from the tondo (left) and Raphael’s Sistine Madonna (right).
as they explain in smithsonian magazineI suspected that the tondo was painted by none other than Renaissance genius Raphael Sanzio. And I used to believe it for one very simple reason: faces Mary and the baby Jesus depicted were similar to the famous Sistine Madonna, a Raphael masterpiece.
The problem is that not everyone shared Winward’s belief in that theory. Others thought the tondo was simply a Victorian copy of The Sistine Madonna, a painting commissioned by Raphael for the church of San Sisto in Piacenza during the second decade of the 16th century by Pope Julius II.
To protect his legacy after his death and to ensure that the secrets of his collection, which of course includes the tendons in question, Winward created the de Bracy Trust Collection in 1995. He had a purpose precisely for continuing the pieces available to scholars … scholars of the arts who are improving their arsenal of resources over time.
An important first step was taken in 2004 when, thanks to Raman spectroscopic analysis, Professor Howell Edwards of the University of Bradford found pigments that were commonly used by artists before the 17th century. What’s more, they identified a vegetable-derived, starch-based gum that pointed to a more specific historical period: the Renaissance.
That discovery – Edwards acknowledged – “contributed to dispel the idea that it was a Victorian copy”, but it remained pending ultimate test To connect Tondo with Raphael. If he had painted it, of course.
Over the years various experts in Raphael’s work have come to conclusions after examining the tondo that reinforce English theory. For example, it is reported that it was painted in Rome. One of them, Murdoch Lotian, went further and suggested that it may predate the Sistine Madonna, for which it may have served as a model. However, for many the final judgment has now come. And not by the hands of scholars, but by artificial intelligence.
Experts from the Universities of Bradford and Nottingham have compared the enigmatic tondo with a painting of The Sistine Madonna, using facial recognition tools that use AI. The aim was to find similarities between the faces of Mary and Jesus in both images and their findings have been spectacular: in the case of the Madonna, the similarity discovered by the AI reached 97% and 86% in case of child. Both percentages are well over 75%, which experts take as a reference to speak of the same level of equality.
“The forensic facial comparison study we conducted confirmed that the faces of the Bracy Madonna and Child and the Sistine Madonna are identical. Viewing the faces with the human eye shows an obvious resemblance, but at the pixel level computers are far more accurate than we are.” can see much deeper,” explains Professor Hasan Ugail from the University of Bradford.
With this data and previous studies, Ugail and his colleagues are blunt: “The same model was used for both images and, without a doubt, they are from the same artist,
To refine the tool, Ugel turned to millions of faces that allowed him to train an algorithm that is responsible for recognizing and comparing facial features across thousands of dimensions. The system uses a deep neural network (DNN) to pass data through multiple filters, allowing it to identify patterns with an accuracy far greater than that of humans.
“The technology can be implemented in a variety of purposesincluding art analysis and even medical care”, highlights the Bradford specialist.
For now, the tool has already managed to put to rest — at least for the study’s authors — an artful debate that has faced scholars for decades.