Researchers at the University of Nottingham and Kingston University in the UK have solved a perplexing riddle that’s been bugging computer scientists for years: how to make a 3D facial reconstruction from just one 2D headshot. The results, on display and ready for you to try out, might be a little rudimentary (and just a tad creepy) but they provide another inventive example of the power of artificial intelligence at work.
Calling it a fundamental computer vision problem of “extraordinary difficulty,” the research team was able to tap into the vast computing power of a Convolutional Neural Network (CNN), a form of artificial intelligence that uses machine learning to “teach” a program through example. In this case, the neural network was fed 80,000 images of faces of both the 2D and 3D variety before it was able to effectively construct its own models from 2D images alone.
While 3D reproduction has been available for some time now — you can get a 3D-printed six-inch high figurine of yourself made by Toronto-based Selftraits, for example — those involve getting photographed from many angles at once in order to cobble together the 3D data set.
Not so for the newly developed technology.
“Current systems often assume the availability of multiple facial images (sometimes from the same subject) as input, and must address a number of methodological challenges such as establishing dense correspondences across large facial poses, expressions, and non-uniform illumination,” say the authors of the new project, to be presented next month at the International Conference on Computer Vision (ICCV) 2017 in Venice.
The fruits of their labours are ready to use now, just by uploading an image and in seconds, receiving a nifty 3D gif, which, depending on your tastes, can be either enlightening or scary (or both). So far, the team reports that the public has uploaded over 460,000 faces since September 7th.
Dr. Vasileios Argyriou from Kingston University’s Faculty of Science, Engineering and Computing and co-contributor to the project said in a statement, “What’s really impressive about this technique is how it has made the process of creating a 3D facial model so simple.”
The technology could have a wide range of applications, from retail services such as make-up and eyewear programs which allow users to try on a variety of looks without leaving the comfort of their couch to better personalized avatars for gaming along with a whole slew of possibilities within the field of augmented reality.
Augmented reality is ready to become more mainstream, says tech columnist Matthew Braga for CBC News. Braga points to the new iPhone 8, iPhone 8 Plus and iPhone X which all come equipped with calibration technology to be able to handle the number of AR apps about to come online.
“Apple is trying to show that augmented reality on a smartphone can do a lot more than what we’ve seen so far from Snapchat or Pokemon Go,” writes Braga. “But it’s too early to say whether these experiences will actually lead to a whole new way of interacting with the world and with our devices, as the tech industry hopes.”