Disney researchers say they have developed a method to help computer-vision systems avoid the confusion associated with changes in perspective.
The new system can estimate the pose of an object, based in part on similarities in how different types of objects appear from the same angle. The method also can use its knowledge of pose to help identify an object, making it useful for a wide variety of computer-vision applications.
The method is based on how people draw inferences from other things they have seen.
"If they know what a bicycle looks like from various angles, that can help them predict what a motorcycle might look like in different poses, because of the visual similarities among these two objects," says Disney researcher Leonid Sigal. For example, a side view of a horse has more in common with the side view of a cow than it does with the front view of a horse. The researchers were able to use these similarities, which are shared across different categories of objects, to develop a metric-learning approach, which forms the basis of the new predictive method.
From Disney Research
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA