Communications of the ACM
Bet You're a Liam--Software Puts Names to Faces
A new system associates names with face with greater accuracy than human guesswork.
Credit: Thinkstock
Cornell University researcher Andrew Gallagher believes it is possible to take an educated guess at someone's name with just one look at their face.
Gallagher's team used tagged photos from the Flickr website to build up a database of named faces and then trained a computer to recognize the contributing factors. For example, Alejandras tend to have darker hair and skin than Heathers.
The researchers tested the system using faces belonging to people with the top 100 names in the United States. Although the computer was only able to guess the correct name about 4 percent of the time, that mark is four times as good as the success rate achieved by random guesses and is about a third better than human guesswork. Gallagher believes accuracy could be improved, and the software could be used to automatically harvest names from online picture captions to tag people in pictures.
Gallagher would like to expand the database to include all known names in the U.S. He plans to present the research at the Computer Vision and Pattern Recognition conference in June.
From New Scientist
View Full Article - May Require Free Registration
Abstracts Copyright © 2013 Information Inc., Bethesda, Maryland, USA
No entries found