Scientists are using billions of photos to analyze fashion around the world
Why do people in Brazil wear more hats than people in Italy? Why do so many Americans wear black? These questions are finally answered, thanks to the wonders of technology.
The clothes, as they say, make the man. But they can also unlock a treasure trove of clues about a person's location, era, culture and socioeconomic status.
It's that wealth of knowledge, combined with the billions of photographs uploaded to the internet every day, that is helping scientists attach artificial intelligence to personal style.
It's all part of a new arm of research from computer scientists at Cornell Tech, a joint venture of New York's Cornell University and Israel's Technion Institute. A few years ago, the two institutions joined forces and won a contest to establish a new applied sciences graduate school in New York City. It's now grown into a campus on Roosevelt Island, between Manhattan and Queens, with futuristic-looking, airy buildings and well-manicured lawns.
The scientists are using what they call "deep learning methods" to detect attributes specific to the location of the photo. "For example, where in the world is wearing hats more common? At what time of the year? Which colors are more popular in summer versus winter?" said Kavita Bala, a computer science professor at Cornell and one of the paper's lead authors. "Our approach produces a first-of-its-kind analysis of global and per-city fashion choices and spatio-temporal trends."
To get started, researchers first weeded out any photos that did not have people in them. Then, they developed an object-recognition algorithm to recognize different articles of clothing and attributes of those items, like sleeve length and pattern.
From there, they could glean tons of data: how people pair their clothing, which trends are more popular in particular places and how styles evolve over time. Like a massive fashion history yearbook.
"The combination of big data, machine learning, computer vision and automated analysis algorithms makes for a very powerful analysis tool in visual discovery of fashion and other areas," said co-author Kevin Matzen, a Cornell graduate student.
MORE FROM THE GRAPEVINE: