Art criticism appears to be a discipline best reserved for humans. After all, who better knows what appeals to humans than humans? On the surface it would seem like a task that we could never automate.
From the Economist we learn of Lior Shamir, a computer scientist from Lawrence Technological University in Michigan. Mr Shamir recently published an article (pdf) in which he suggests computers may have as good an eye for style as humans.
Dr Shamir’s team scanned fifty-seven works by nine artists into a program which assessed descriptors based on Fisher scores and the most informative features were used to classify the paintings by artist, similarities and schools of art. The program was able to identify the artist with 77% accuracy and it correctly identified the school of art 91% of the time.
To look for comparisons between artists, the team programmed a statistical method that scores the values of descriptors between artists. As a result, the computer was able to see similarities that have escaped humans:
Surprisingly, the values of 19 of the 20 most informative descriptors showed dramatically higher similarities between Van Gogh (left above) and Pollock (right) than between Van Gogh and painters such as Monet and Renoir, who conventional art criticism would think more closely related to Van Gogh’s oeuvre than Pollock’s is. (Dalí and Ernst, by contrast, were farther apart then expected.)
Whether Pollock was actually influenced by van Gogh or stumbled upon those similarities by chance remains to be seen. As the Economist notes, it provides art historians a new line to explore.