NTechLab is only one year old, but it’s already making headlines and playing with the big boys in the industry. They won the “MegaFace” facial recognition competition held last year in Washington state using a controversial technology, ahead of Google and Facebook’s own algorithms.
Face recognition has grown massively in past years. It was typically used in security systems but more recently has seen commercial applications in social media and marketing. Simply put, there’s a lot of money in facial recognition nowadays and tech giants Google and Facebook want to be ahead of everyone’s game. But a scrappy Russian startup came as a surprise contender.
“We are the first to learn how to efficiently handle large picture databases,” said NTechLab founder Artem Kukharenko to Intel iQ. “This advantage is the key to solving real-world problems, such as finding a criminal in real-time or identifying a regular customer from store surveillance cameras.”
NTechLab’s tech uses deep learning and a neural network-based architecture – which means that it learns how to identify faces and gets better all the time. According to Kukharenko, the most complicated step is the first one, in which the algorithm initially recognizes how to identify a face. After it does that, it creates a vector of 80 variables which store detailed information about the face. The algorithm differentiates between changeable parameters (facial hair, wrinkles, etc) and stable parameters. Then, it checks in those variables and makes the identification. It’s a stunningly fast process.
“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.”
So the 20-people company seems to be set for glory and there is already talk of some massive contracts being discussed for them. But in the meantime, if you want to test their software’s capabilities, you can try out FindFace, a free app which basically works like Shazam for people. You snap or upload a photo and then the app tries to identify the person by searching through the Russian social network Vkontakte.