In an unsuspecting business park in Moscow, facial recognition is reaching new heights. It’s not from a division of Google or Microsoft, or by a covert agency, but instead by a small startup that currently employees less than ten people. The startup is NtechLab and they are changing the face of facial recognition.
Artem Kukharenko, one of the founders of NtechLab, has always had a passion for math - the backbone to almost any computer program, and especially pertinent in something like facial recognition. After attending one of the most mathematically demanding grade schools and winning an all-Russia Information Technologies Olympiad, Artem went on to Moscow State University where his interest in Computer Vision and Computational Mathematics continued to grow. After graduation, Artem took his talents on the road, living in Argentina while freelancing for American school, Purdue University, in Indiana. His job was to create algorithms to identify objects on video - one application being self-driving cars and recognition of of road signs and obstructions.
After his time in Argentina, he returned to Moscow to work at Samsung, where he continued to worked in the field of neural networks. It was during this time (he had a lot of free time one New Year’s holiday) that NtechLab and eventually, FindFace, would find their start. With the help of his girlfriend, Artem created an app that would determine a dog’s breed by uploading one photo. The app had mixed success, with many people complaining of the accuracy, but it was just the start he needed to motivate himself even further.
Thanks to a friend, Artem was able to show his app to venture fund, Typhoon Digital Development,” and the rest, as they say, is history. Artem received funding, and NtechLab was formed. NtechLab’s power comes in the form of its advanced algorithm and neural networks that actually learn as they absorb more data. It can be used in a limitless number of applications and services to provide facial recognition and matching. While uses are currently limited, there are some implementations already being used.
FindFace is the most notable of services currently using NtechLab’s powerful algorithms. The service uses the Russian social network, Vkontakte, to search through faces that the software has scanned to find positive matches. Maybe an old friend or doppelgangers of your favorite celebrity. The success rate is pretty impressive, as well, coming in at over 70%. The FindFace app has been downloaded over 1 million times and all users have to do is upload a picture into the app and FindFace gets to work. The app will scour through approximately 1 billion profile photos in a matter of seconds and return with who it believes it to be, as well as ten other possibilities.
The facial recognition also made its way to a Russian rave this year. The Alfa Future People festival purchased the technology to offer festival-goers a way to find pictures of themselves from the festival. They simply had to upload a selfie into one of the kiosks and FindFace would find other pictures of them from the giant party.
And it’s not limited to Russian social networks, either, the underlying technology can be implemented into multiple applications. One example, think loyalty cards. Going into a store and the system picks up on you entering, seeing your more recent purchases and applying discounts accordingly. People you met at parties or on vacation could be found again by uploading any pictures you took together. Casinos could use it to see when big spenders come in the doors or when someone who was banned struts in. Law enforcement could use it to identify suspects that were picked up on security cameras during a break-in or robbery. Missing children, the elderly, or anyone could be found quickly thanks to this technology. The social implications are huge.
Other companies have facial recognition software, but at present it seems that FindFace is the most advanced out of all the options. At the University of Washington ‘Megaface’ competition, the Russian startup beat out Google and a program from China in a category that compared largest number of images to accuracy. FindFace finished with 73.6% accuracy, while Google and China finished with 70.5% and 65.2%, respectively. Not bad for a small team out of Russia. Sadly, Facebook opted out of the competition, it would have been interested to see how FindFace compared to them.
While people will have their concerns about the software regardless, it seems if it’s only a matter of time before we start seeing this in our daily lives. Eventually ads in stores may be catered to us based on our preferences. The possibilities truly are endless. In an interview with Bloomberg, co-founder Alexander Kabakov says, “In theory, our technology can index 5 billion people - almost the entire population of the world.”
