Selfies are a controversial new phenomenon with the advent of affordable and high quality camera phones. Some think of selfies as the new self portrait, being able to have full creative control on how they present themselves to the world. But others see it differently, as a vain and childish way to present your identity. Because there are so many strong feelings, it’s hard to objectively analyze selfies. Even with the aid of computers, it is very difficult to parse these images effectively. That’s what Lev Manovich wants to fix.
Last Friday, Manovich, a professor of computer science working at the University of California San Diego and City University of New York, gave a talk at EMPAC Studio B relating to his new website, Selfiecity. This new website is the results of his analysis of selfies from Instagram photos taken in five major cities around the world. From each city, tens of thousands of images are taken, then sent to a program on Amazon known as Mechanical Turk. This program gets people to do human intelligence tasks—menial labor that a computer cannot analyze—usually for large data sets that a single person or small group cannot parse in a reasonable time frame. Of the thousands of photos taken, only about four percent were selfies. Next, the
Mechanical Turk workers estimate the age and gender of each person in the selfies. Finally, facial recognition software determines whether the subject has glasses, how much they tilt their head, and their mood. From this, Manovich and his team got 640 photos from each city, a good sample size for examining selfies and trends in each city.
On the website, they posted their findings as well as an interactive menu where you can isolate certain selfies by the city, gender of the subject, amount of head tilt, or other factors. The findings are quite interesting: the gender disparity in Moscow, how much people from Bangkok, it’s strange to be looking at trends in something as new and strange as selfies.
But Manovich doesn’t want to stop with just selfies. He and his team intend to apply this to art like self portraits and find the same trends through different art periods and locations. There’s a lot of potential in this form of data analysis, and Manovich intends to take advantage of it. For more information on this project, visit http://selfiecity.net/.