Years ago, I watched an instructional video all about how to remember names. The video had been directed towards businessmen and salespeople who might have to recall the names of a host of people during their career, some people are talented at connecting faces to names—others, like me, do it poorly. Now, thanks to Viewdle, your smartphone shall come to the rescue and automatically recognize people for you.
The software would essentially attempt to smart-tag a users photos before they uploaded them to sites like Facebook by scanning the images, discovering faces, and then using facial recognition to identify the people in them. We saw them receive $10 million back in October in investment funding, so when they appeared at the Mobile World Congress and started showing things off means they’re actually going somewhere.
Viewdle sits between the camera and the user analyzing faces in the camera stream, identifying them, then offering links to Facebook, YouTube, LinkedIn, and other social media platforms. A user can identify and tag people in pictures & videos then pass the information to their social networks. As they tag others the software learns to recognize them, and can even share these new visual profiles with other users. The live view also offers an augmented reality tagging overlay that reveals information about the people around you.
According to Viewdle themselves, they intend to cast a wide variety of implementations but the only one currently available happens to be a desktop program called the Viewdle uploader. However, the quote above taken from their website suggests that they intend to cast a series of different implementations including an augmented-reality version that will automatically tag faces seen by a smartphone’s camera.
According to sources, the Viewdle mobile apps will set themselves apart from other facial recognition software because it will do all of its processing on the local device. Perhaps this will be a good app to leverage the sheer processing power being put into the industry by nVidia chips for mobile devices. From a code perspective, the facial recognition will probably use some type of simple facial geometry recognition in order to pick points and ratios on the apparent face and then save those into the phone’s memory. From the looks of it, the Viewdle system does not link into the cloud for either storage or processing; but does reverse-process tagged photos from social websites in order to start identifying people themselves.
The next evolution of this sort of software, and the really creepy, “the Internet is stalking you” direction it could go would be a service that allows other people to identify you by processing already tagged photos and making guesses. However, Viewdle seems to be entirely user oriented and doesn’t reach outside of your network for tagging information—in fact, the local processing and storage would severely limit its stalker-esque utility.
We can probably expect to see a lot more apps like Viewdle attempting to capture the social media zeitgeist and part of that reason is that Facebook has been pushing hard for facial recognition. Of course, the funding led to a Facebook facial-recognition feature called “Tag Suggestions” in December, however, almost nobody uses it. While facial recognition is fun, because it doesn’t extend to people’s mobile devices and act as another augment, few people actually hook into it except for its novelty.