Communications of the ACM
Not Ok, Glass
A new software can be trained by users to automatically identify places where devices like Google Glass should not capture images.
Credit: GlassKap
Indiana University researchers have developed PlaceAvoider, software that uses computer-vision techniques to automatically identify potentially confidential or embarrassing pictures taken with new devices such as Google Glass.
PlaceAvoider users train the software by taking pictures of the rooms they want to blacklist. The software then flags new pictures taken in those rooms so the user can review them.
The program uses an existing algorithm called scale-invariant feature transform to identify regions of high contrast around corners and edges within the training images that are likely to stay visually constant in varying light conditions and from different perspectives. For each of these factors, the algorithm produces a numerical fingerprint consisting of 128 separate numbers relating to properties such as color and texture.
During testing, the system accurately determined whether images from streams were from blacklisted rooms nearly 90 percent of the time.
"Algorithms could be used to automatically organize these huge collections of images to make them safer, more browseable, searchable, and useful," says Indiana University professor David Crandall.
From Technology Review
View Full Article
Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA
No entries found