Pages Menu
Categories Menu

Posted by on Oct 2, 2012 in Technology |

Computers also are art experts

Computers also are art experts

Computers are now able to assign paints artistic styles. They can also recognize sketches drawn by people. Another task that exceeds the Turing test.

In a previous article I applied my particular formulation of the Turing Test for tasks :

A computer exceeds the Turing Test for a task X when a human judge can not distinguish between a human and a contestant contestant computer running the task X

A group of Lawrence University in Michigan has developed a program that is able to analyze various tables of universal history and group them according to their similarities in styles. The program performs the task better than untrained humans.

1000 frames were exposed from 34 well-known artists. The result goes well with the perception of experts.

Tables grouped into two groups, classic and modern. Within these groups was able to find similarities between Gauguin and Cezanne, Post-Impressionists, including Salvador Dali surrealist, Max Ernst and Giorgio de Chirico, between representatives of the High Renaissance as Raphael, Leonardo Da Vinci and Michelangelo or Vermeer Baroque as , Rubens and Rembrandt.

The program studied up to 4,000 parameters describing characteristics of the paintings as texture, color or shape. Overall it is an excellent example of the fields that are being applied AI Artificial Intelligence.

Computers also are art experts image 2

Another program , this time developed by the universities of Berlin and Brown (Providence, USA), is able to recognize sketches and find that object is.

250 categories were established and used 20,000 sketches. The program was able to make the correct identification in 56% of cases, slightly below humans who achieved a 73% success rate.

Using crowsourcing recruited people who were drawing the sketches. The sw is able to recognize objects as being drawn. There is even a version for iPhone called WhatsMySketch .

The data collection is the only way to get the machine learning. There is no way to learn to recognize sketches of lions based on an intelligent algorithm. The algorithm needs to see nearly 100 examples of how people draw the lions to be able to distinguish these animals from potted plants.

As more and more often, sw may use the user experience. Just as Google uses your activity to refine their algorithms (search, voice recognition …) the new sw can use the input from users to improve their effectiveness.

The visual pattern recognition is no longer an exclusive power of humans whether it is drawn sketch of a rabbit like a Goya painting.

Tags: ,