Communications of the ACM
Researchers Develop "visual Turing Test"
Athens, Baltimore, Hong Kong, Miami: what are those people doing? A new evaluation method measures a computers ability to decipher movements, relationships, and implied intent from images by asking questions.
Credit: News from Brown
Researchers at Brown and Johns Hopkins universities have developed a "visual Turing Test" that can evaluate how well computers understand information taken from images.
The system is designed to test for a contextual understanding of photos and works by generating a series of yes or no questions about an image, which are posed to the system being tested. Each question is progressively more in-depth and based on the responses to the questions that have already been asked. The questions are geared toward measuring the computer's understanding of the contextual "storyline" of the photo.
"You can build this notion of a storyline about an image by the order in which the questions are explored," says Brown professor Stuart Geman.
The system is more objective than having a person ask a computer about an image because the questions themselves are computer-generated. However, a human is required to tell the test system when a question is unanswerable because of the ambiguities of the photo.
Geman says the test could lead to new ways of teaching computers how to look at images.
From News from Brown
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found