Viewdle brings automatic photo tagging to smart phones
There's something Orwellian about machines that can identify people by looking at their faces. In fact, the facial-recognition software being developed by Viewdle, a San Francisco-based start-up, has its roots in technology created for the surveillance-happy government of the former Soviet Union. But Viewdle has found broader and far more consumer-friendly uses for its wares.
The company has been conducting an invitation-only trial of desktop-computer software that automatically identifies faces in photos and videos, then "tags" the files with this information. At the Consumer Electronics Show last week, Viewdle demonstrated an application that can do the same with a smart phone camera, automatically and in real time. That made Viewdle the first company to show "real visual computing that works in a mobile device," Chief Executive Laurent Gil claimed.
The company, which is selling the technology to other companies in the mobile-phone industry, expects versions of the app to be available by the end of the year.
Viewdle automates the labor-intensive process of tagging photos and uploading them to Facebook. That sort of convenience will appeal most to young people who are "all about sharing the moment" and documenting their lives through their mobile-phone cameras, Gil said.
The app determines identities by comparing the faces seen by the camera's lens with images previously stored on the phone. That's a twist on the usual practice of trying to match faces against a vast online database. It also uses a comparison technique -- morphing images to try to match unknown faces with known ones -- that's faster and requires less computing horsepower than the conventional practice of comparing multiple features of the unknown face against those in identified images, Gil said.
For the app to work, users have to build up a database on their phones of tagged photos that Viewdle can use as a reference. The software makes the task easier, though, by logging into Facebook and collecting information from the photos they've tagged there.
Facial recognition is just the first step for Viewdle, whose morphing technique can be used to recognize any physical object seen by a camera -- stairways, cars, animals, you name it. Other potential uses include augmented-reality applications for gaming and commerce, said Jason Mitura, Viewdle's chief product officer. For example, a retailer could add virtual displays to its store stocked with its online-only inventory, with interactive 3-D images of those products visible through a smart phone camera lens.
-- Jon Healey