Home > Cool Tech > This scrappy Russian startup is beating Google and…

This scrappy Russian startup is beating Google and Facebook at facial recognition

Whether it’s facial recognition on Apple’s iOS 10 or Facebook learning to recognize human faces virtually as well as another person, there’s no doubting that facial recognition technology is big business right now.

But while both of those companies have massive, multi-billion dollar budgets behind them, a small Moscow-based artificial intelligence startup named NTechLab may have stumbled upon one of the best facial recognition systems around.

How good? At last year’s “MegaFace” facial recognition competition in Washington, it managed to best a number of rivals — including Google’s own FaceNet. It’s since been turned into an app called FindFace, which has been downloaded 600,000 times in Russia since February, while roughly 300 inquiries have been made by companies and governments around the world interested in the company’s underlying technology.

unnamed2

“We have found a special type of internal architecture for neural networks, that perfectly fits the face recognition tasks,” NTechLab CEO and founder Artem Kukharenko tells Digital Trends. “To search among huge datasets — up to billions [of images] — we use our specially developed search engine, which is extremely quick and accurate. Each face in the search index is represented by only 80 numbers (a very small amount for such algorithms), and the overall search time is [only] about half a second.”

According to Kukharenko, FindFace has so far performed about one quadrillion photo comparisons using images from the Russian social network Vkontakte, which has around 200 million profiles. The dream of FindFace is quite literally to be able to do what its name suggests: to let users see someone on the street, snap a quick photo, and immediately be able to link that person to their social media profile.

unnamed

Of course, such applications immediately raise privacy concerns — just as similar fears did about the ill-fated Google Glass. But Kukharenko argues that his technology, carefully moderated, can be useful. “We will thoroughly monitor how our partners are going to use our platform to make sure it will be used appropriately,” he says, discussing with Digital Trends his plans for a new cloud face recognition software platform, which will be available for businesses to plug into and use for their own recognition tasks. He also brings up use-cases such as how FindFace has already helped track down criminals.

“Two pyromaniacs in Saint Petersburg arranged the arson of a building, and the camera record was the only clue,” he continues. “Someone uploaded the video footage to Pikabu, a Russian online community, which is quite similar to Reddit, and asked for help. Soon the commentators suggested using FindFace to find the criminals by taking frames from the [camera footage]. FindFace successfully identified those people, and with that information the residents were able to inform the police, who later found them.”

There may be a few ongoing questions about facial recognition, but one thing is for certain: tools like FindFace are here to stay, and getting more accurate all the time.