Art experts could soon be replaced by computers as scientists of Lawrence Technological University in Detroit have developed software that can identify, evaluate and attribute works of art.
Computer scientists Lior Shamir and Jane Tarakhovsky created the software that focuses on 4,000 numerical image descriptors and analyzes form, texture, and visual content of the paintings without any human guidance.
Their work has been published in the ACM Journal on Computing and Cultural Heritage. The program has managed to precisely attribute around 1000 paintings of 18 modern and 16 classical painters with no mistakes.
The computer automatically divided the 34 well-known painters into groups showing that it s able to identify painters of the same artistic movements.
It placed the High Renaissance artists Raphael, Leonardo Da Vinci, and Michelangelo close to each other. Then separated the Baroque painters like Vermeer, Rubens and Rembrandt. Van Dyke, Durer and Bruegel were united into another group. Similarly it separated Gauguin and Cézanne and united Salvador Dali, Max Ernst, and Giorgio de Chirico into one group.
A similar attempt to create a program that would be able to analyze visual content has been showcased by Google. That software was good enough to be able to find cats in various images.