Israeli computer program can ID artistic painting style

Van Gogh’s ‘Irises’ – ‘As soon as the computer learns to recognize the swirls of Van Gogh, it will recognize them in pictures it has never seen before.’Art aficionados aren’t normally known for their resemblance to computer scientists – one …

Van Gogh’s ‘Irises’ – ‘As soon as the computer learns to recognize the swirls of Van Gogh, it will recognize them in pictures it has never seen before.’Art aficionados aren’t normally known for their resemblance to computer scientists – one might reasonably expect the hush of a climate-controlled gallery and the electronic buzz of a software lab to be mutually exclusive habitats.

But one innovative program developed at the University of Haifa is proving that assumption to be rather – well – artless, exploring a gray area between mathematics and fine arts which, incidentally, is anything but gray.

According to Professor Daniel Keren of the university’s department of Computer Sciences, the key is in defining the somewhat nebulous notion of ‘style’. By transforming images of nature, people, flowers and other scenes to a series of mathematical symbols, sines and cosines, his team has developed software which ‘teaches’ a computer to identify the unique visual signatures of various artists.

After being primed in such a way, the program is able to identify a particular artist’s ‘style’ in other works – even if they are works which have not been fed into the system previously.

“As soon as the computer learns to recognize the clock drawings of Dali, it will recognize his other paintings, even without clocks. As soon as the computer learns to recognize the swirls of Van Gogh, it will recognize them in pictures it has never seen before,” says Keren.

But is it really about art? One might object that the meaning of a painting has nothing to do with mathematics – and Keren is quick to point out that his software isn’t about to put critics out of work anytime soon.

“I think that this captures a painting in a very basic way. It samples very small patches, at the micro-level. It can’t yet look at a painting as a whole, and say, this is characteristic of a certain artist because it represents a particular subject – a cow, or a flower, for example.”

But while the technology might not yet put art experts out of work, it may prove helpful for amateur enthusiasts and collectors. One application might be in the area of forensic analysis.

Keren’s team has already fielded enquiries from art dealers and auction houses, who deal with the problem of forgeries frequently. If the software is able to assist in the identification of a forged Van Gogh or Matisse, collectors and galleries may find the simple computer program invaluable to tell them if a purchase is wise – or financially foolish.

Whether this will prove enough to bridge the divide between gallery and lab remains, however, another question entirely. “It will be a long time before we have a computer art critic,” laughs Keren. “Maybe one day… who knows.”


*/