Researchers applied AI analysis to one of Raphael’s most famous paintings, the Madonna della Rosa.
This AI, documented in a study published in Heritage Science, has identified that the face of St Joseph in the artwork was likely not wholly painted by Raphael himself.
That doesn’t mean it’s a forgery or counterfeit in any way, however. It was common for Renaissance artists to collaborate with students and assistants, which has created debate about correctly attributing works from this era.
While confirming an artist’s contribution to a painting typically requires largely subjective analysis, the researchers, leveraging a machine learning algorithm, discerned different techniques used in the work to peel back its origins.
The researchers used a well-known AI model called ResNet50, originally designed for recognizing objects in images. They adapted this model to focus on artworks, especially to identify features unique to Raphael’s paintings. This allowed the AI to learn and recognize Raphael’s distinct styles and techniques, which are not always obvious to the human eye.
Alongside ResNet50, the study used a method known as Support Vector Machine (SVM). This technique helped the AI to classify the artworks more accurately – essentially, to decide whether a painting was likely done by Raphael or not.
This combined approach proved highly accurate, with a 98% success rate in identifying Raphael’s paintings.
Results for a section of The Sistine Madonna – an authenticated painting by Raphael. Source: Heritage Science.
Researchers further developed techniques to analyze the brushstrokes in a painting. Since each artist has a unique way of applying paint, this analysis provides another layer of verifying authenticity.
Mathematician and computer scientist Hassan Ugail from the University of Bradford explained the process: “Using deep feature analysis, we used pictures of authenticated Raphael paintings to train the computer to recognize his style to a very detailed degree, from the brushstrokes, the color palette, the shading and every aspect of the work,” he said.
Ugail emphasized the AI’s extended capability versus human analysis, noting, “The computer sees far more deeply than the human eye, to microscopic level.”
The analysis revealed that while the Madonna, Child, and St John were consistent with Raphael’s work, St Joseph’s face did not match his style. “When we tested the della Rosa as a whole, the results were not conclusive,” Ugail remarked.
“So, then we tested the individual parts and while the rest of the picture was confirmed as Raphael, Joseph’s face came up as most likely not Raphael.”
The AI’s findings add weight to the long-standing theory that Giulio Romano, a pupil of Raphael, may have contributed to the painting, particularly in the depiction of St Joseph. However, this remains unconfirmed.
Ugail described how AI forms just one part of a multifaceted authentication process that “involves looking at many aspects, from its provenance, pigments, condition of the work, and so on. However, this sort of software can be used as one tool to assist in the process.”
Not long ago, researchers created an AI model capable of discerning the origin of wine based on its “terroir,” which is used to describe an elusively complex blend of chemical signatures.
Now, the technology is identifying artwork’s authenticity, illustrating how ambiguous and highly subjective topics can be broken down through novel machine learning.
The post Researchers use machine learning to analyze artwork authenticity appeared first on DailyAI.
