Quantcast

Computer Algorithm May Soon Be Picking Hipsters Out Of The Crowd

December 11, 2013
Image Caption: Examples of social groups in the researchers' Urban Tribes dataset. Credit: UCSD Jacobs/ Urban Tribes Project

Lee Rannals for redOrbit.com – Your Universe Online

Computer scientists are developing an algorithm to help end the debate on whether or not someone is a hipster.

Calling someone a hipster can be either an insult or a compliment depending on the person who is on the receiving end. The term has had a very broad definition and has been the subject of debate, with some claiming it to be a clothing style and others a lifestyle. However, computer scientists at the University of California, San Diego are seeking to end the debate with an algorithm that determines whether someone is a hipster, a surfer or a biker.

So far, the algorithm has been able to correctly identify a social group 48 percent of the time, which the researchers say is actually a very good number.

“This is a first step,” Serge Belongie, a computer science professor at the Jacobs School of Engineering at UCSD and co-author of the study, said in a statement. “We are scratching the surface to figure out what the signals are.”

The researchers hoped their algorithm could make it easier to pick up social cues, like clothing and hairdos, to determine people’s urban tribes based on visuals featuring more than one person. This technology could be used in applications like generating more relevant search results and ads.

The algorithm segments each person in six sections, including face, head, top of head, neck, torso and arms. This method is an example of what is better known as a “parts and attributes” approach. The team designed the algorithm to analyze the picture as the sum of its parts and attributes.

Researchers fed the algorithm pictures labeled for the urban tribes they represent, and then fed the algorithm pictures without labels. The computer program was able to predict a social group more accurately than just a random setting. The next step for the team is to run the same set of pictures by human users and see how they perform and compare the two results.

The computer scientists said they referred to Wikipedia to help define the urban tribes. They selected the eight most popular categories in the online encyclopedia’s list of subcultures, which also included country, Goth, heavy metal, hip hop and raver. The team gathered photographs from three common categories for social venues, such as formal events, dance clubs and casual pubs.

The team said they plan to make the collection of photographs gathered available to other research groups who are interested in studying urban tribes.


Source: Lee Rannals for redOrbit.com - Your Universe Online



comments powered by Disqus