05:36 AM

Predictive Computer Vision: The Face Of The Future Looks Wrinkled


Greg Leeming is the director of the Intel Technology Science Center for Visual Computing at the University of Washington.

By Intel Free Press

Have you ever wondered what you'll look like in 10, 20 years? The University of Washington and Intel Labs has embarked on a facial aging project using big data to analyze and predict the way people's faces age.

Demonstrated at Intel Developer Forum 2014, finding your predicted future face is exceptionally easy through the use of an iPhone app. All the user has to do is input some information relating to age, gender and ethnicity, then select or take picture of herself using the front-facing camera. The rest then appears like magic.

Behind the scenes, however, are big data and clever algorithms. Greg Leeming, director of the Intel Technology Science Center for Visual Computing at the University of Washington, explained that behind the curtain is a big data approach.

"We go out to the Internet and grab huge numbers of pictures and organize those pictures into collections of flows. For an individual flow, we would have pictures when a person was a baby, age 3, age 4, etc., and create sequences of how those people age over time," Leeming explained.

"We then apply computer vision to those sequences of pictures to track how pixels move in an optical flow pattern from one picture to the next, and with that you can predict how people are going to age over time."

"Once you average many of those sequences together, you get a sense of how people age. And it turns out people age in very predictable ways, and so now one can apply the algorithm to an individual's picture and show how that person is going to age very, very accurately."

The research team, Ira Kemelmacher-Shlizerman, Supasorn Suwajanakorn and Steven M. Seitz of the University of Washington, released a research paper on this facial aging technology.

Besides serving as a looking glass into our future selves, the facial aging algorithms could prove to be an invaluable tool to supplement law enforcement sketch artists in the long-term search for suspects or missing children.

Foremski's Take: I can't imagine many, or any people want to see how they look in the future -- it's not going to be a pretty sight. But what if the app showed you how you would look if you lost 20 pounds? Or if you gave up smoking and drinking? Or you did yoga/gym more often?

Unfortunately,  there's no app that can stop the march of time.