Computer scientists at the University of California, San Diego (UCSD) are developing a computer vision algorithm that uses group pictures to determine which social group, or urban tribe, a person may belong.
The algorithm, which is currently 48% accurate on average, classifies people into the following groups: Formal, hipster, surfer, biker, or goth, according to a UCSD press release. In the future, the team hopes that the algorithm may be used to generate more relevant search results and ads, allowing social networks to provide better recommendations and content. But its potential usefulness would not end there, as the algorithm could also be used in surveillance cameras in public places to identify groups rather than individuals. In addition, an algorithm that could be used to identify and separate people according to which social group they belong to, would likely be useful in a machine vision setting such as pattern matching.
A 48% accuracy of the algorithm is considered a first step by UCSD computer science professor Serge Belongie, who says that the team is "scratching the surface to figure out what the signals are." Still, the accuracy rate can be considered a good result, considering that the algorithm is able to look at group pictures rather than pictures of individuals, which they hope will enable them to pick up certain cues to determine the "tribe" of an individual based on visuals featuring more than one person according to the press release.
The algorithm segments a person into six sections: face, head, top of the head, neck, torso, and arms. Essentially, the computer vision algorithm analyzes a picture as a sum of its parts and attributes. The ultimate goal is to improve the analysis of facial features and other attributes within the system to eventually improve upon the 48% accuracy.
View the UCSD press release.
Share your vision-related news by contacting James Carroll, Senior Web Editor, Vision Systems Design
To receive news like this in your inbox, click here.