Sign In

Communications of the ACM

ACM TechNews

Computers Predict People's Tastes in Art

View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
Part of an abtract painting.

A new California Institute of Technology study shows that a simple computer program can accurately predict which paintings a person will like.

Credit: Joseph Schillinger

Researchers at the California Institute of Technology (Caltech) used a program to predict people's art preferences.

The team recruited more than 1,500 volunteers via Amazon's Mechanical Turk crowdsourcing platform to rate paintings in various genres and color fields, then fed this data to the program.

The researchers taught the computer to deconstruct a painting's visual properties into low-level features (contrast, saturation, and hue) and high-level features that require human evaluation.

Caltech's Kiyohito Iigaya said the program combines these features to calculate how much a specific feature is accounted for when deciding on the artwork's appeal; afterwards, the computer can accurately forecast a person's preference for a previously unseen work of art.

Caltech's John O'Doherty said the research reveals insights about the underpinnings of human aesthetic judgments, "that people appear to use elementary image features and combine over them."

From California Institute of Technology
View Full Article


Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account