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