With $10 Million Series B Round, Viewdle Turns Its Face Recognition Technology on Consumers
Viewdle is one of those technology startups that probably should have stayed in stealth mode a lot longer. The San Jose, CA-based company created a stir around the time of its public launch back in September 2007 with demos that promised real-time facial recognition for digital video—a still-unsolved problem that, if definitively licked, could help to make Web video just as easy to search as text. Building on computer-vision algorithms first developed for the Soviet military by Ukrainian computer scientist Mikhail Schlesinger, the company said it would help large media organizations index and better monetize their video archives—claims that propelled the company to first place in the LeWeb startup competition in Paris in 2008.
But then Viewdle did a slow dissolve, so to speak. Three years after its public debut, the startup still hasn’t brought any products to market, and its website, which once featured examples of its facial recognition software detecting the faces of celebrities like George Clooney and Britney Spears, now shows just a logo and a “Stay Tuned!” notice. So it wouldn’t have been unreasonable for an outsider to conclude that Viewdle was no longer really on the air.
Except that it is. In July, John Albright, co-managing partner of RIM’s BlackBerry Partners Fund, told Beet.TV that the fund had put some Series B money into the startup. And today the company revealed more details about the funding round, which totals $10 million and also includes Best Buy, Qualcomm, and existing investor Anthem Venture Partners, which put an undisclosed amount of Series A funding into the company in 2007. The startup will be launching “exciting new consumer applications” within the next few months, CEO Laurent Gil said in the announcement.
The company isn’t talking about exactly what those applications might be, but chief product officer Jason Mitura told me that the strength of Viewdle’s latest facial recognition technology is that it can run on many platforms, including smartphones. “One of the differentiators of our technology is that it works everywhere from the palm of your hand to the cloud, for pictures and videos,” Mitura says. He says the company’s philosophy is that the billions of digital photos and videos that consumers capture every year would be much more useful if they could be automatically tagged—at the moment they’re captured, shared, or uploaded—with the names of the people who appear in them.
Reasoning backward from these clues, and from the identities of the Series B investors, it seems safe to say that Viewdle’s emphasis has shifted from the media and publishing world to the consumer market, and that it’s developing facial recognition software that could be embedded in consumer gadgets such as smartphones, digital cameras, and videocams.
“For consumers, it’s all about real time,” said Kuk Yi, vice president of Best Buy’s venture capital wing, in today’s press announcement. “Viewdle is leading the market by creating compelling consumer experiences that are both real-time and cross-platform—that is why we invested in the company.” Yi has joined the Viewdle’s board as a result of the investment, as has Albright.
Viewdle has an exclusive commercial license to facial recognition technology that chief scientist Schlesinger, a computer vision expert at the Institute for Computing and Information Technologies in Kiev, Ukraine, originally developed for defense and security applications. The company still has 45 engineers in Kiev, including eight PhDs, Mitura says.
While there’s been endless hype about facial recognition since the terrorist attacks of 2001, the technologies haven’t really started to deliver until recently, Mitura acknowledges. “People have lost faith in it, but it’s getting to the point now that it works even in uncontrolled environments, with different lighting angles,” he says. “Professor Schlesinger’s breakthrough was a matching algorithm that did a very good job of normalizing, and making the core pattern recognition robust enough that it worked both in high-quality and low-quality environments.”
The technology has gained both in precision—its ability to correctly identify a face in a photograph or a frame of video—and consistency, its ability to pick out the same person in a series of images. That could be a boon for people who upload lots of pictures and videos to Facebook and other photo sharing sites. Currently, it’s too much work to identify and tag every person in every photo manually, so most people don’t bother. But if this happened automatically, at the moment a picture was snapped, it could lead to much wider sharing and online conversation. “Automatic tagging solutions fit the new dynamic prevalent on the Web, where you want stuff to get posted as its being taken, and you want the people in it to receive a notification as it’s being posted,” says Mitura.
In 2009, both Apple and Google added face recognition capabilities to their popular photo management apps, iPhoto and Picasa. But Mitura points out that these features only work on the desktop and laptop versions of the applications, whereas Viewdle’s algorithms require less processing power—meaning they can run on the same smartphones, digital cameras, or videocams used to capture the images.
Will Viewdle’s new technologies be worth the wait? We’ll know by early next year, when the company expects to make some actual product announcements. “It’s taken a long time and a lot of people and investment to get to the place we are right now,” Mitura told me. “We’re on a shift from being very internally focused on research to being externally focused on products. If you call in six months, the balance will not be as out of whack. And if you look at the investors, all of them have in mind getting this technology into the palms of consumers.”