Artificial intelligence (AI) is actively used today to create images that are difficult to distinguish from real photos or paintings by famous artists. But a group of scientists from Great Britain and the United States has developed a method using AI that allows you to find unusual things in real paintings of classics of fine art.
The study, the results of which were published, examines the canvas “Madonna with a Rose”. Art historians have long debated how fair it is to attribute this painting to Raphael. New findings made by AI add arguments to those who see the work of other artists in the picture, informs with reference to the magazine Heritage Science
As the researchers explained, they “fed” the computer program images of those paintings that are definitely known to belong to Raphael's brush. The program received enough information about the style of this artist – the nature of strokes, color palette, shading, etc.
“The computer sees much deeper than the human eye, down to the microscopic level,” the authors of the study explain.
Machine learning processes, as a rule, need to be trained on a huge number of examples, which is not always possible when it comes to the work of a single artist. In this case, the team modified a pre-trained architecture developed by Microscoft called ResNet50, combined with a traditional machine learning technique called Support Vector Machine.
After studying the painting “Madonna with a Rose”, the program came to the conclusion that the images of the Madonna, the Child and St. John were created by the hand of Raphael. But not the image of Saint Joseph. Researchers note that in previous debates about the painting's authenticity, it was believed that the face of St. Joseph was executed worse than others in the frame.
Giulio Romano, one of Raphael's students, may have been the one who painted the fourth face, but this is by no means certain.