Russian Computer Vision Company is World Leader
Russia seems to get a bad rap these days when in fact it’s a fascinating place to visit. Say what you will about Russia’s President Putin, but he’s managed to turn Moscow into a great city in which every Russian wants to live – and you probably would too after spending a few weeks there. We recently airdropped one of our MBAs into the center of Russia’s capital city to collude with tall leggy blondes, eat pickled fish, and find some Russian startups with interesting stories to tell. One such startup is NtechLab, a company that made our list of the Top 10 AI startups in Russia and also our list of the Best Facial Recognition Algorithms. We popped into their Moscow office to see just what all the fuss is about.
Founded in 2015, Russian startup NtechLab has taken in $1.5 million in funding to develop a client-side facial recognition solution where no biometric data is transferred or stored by the company. They were also one of the companies that made our list of The Top-10 Russian Artificial Intelligence Startups. Their service not only verifies or identifies faces, but recognizes age, gender, and emotions. The company’s algorithms received praise in Washington in 2017, winning the first-ever facial recognition competition devised by the R&D team run by the Director of National Intelligence. With 2,000 customers globally including the UK, US, and China, NtechLab’s technology is used in public safety, dating, security, banking, retail, entertainment, and events organization.
We sat down to talk with NtechLab’s CEO, Alex Minin, to learn about how NtechLab is going head to head with Chinese computer vision startups that have taken in hundreds of millions in funding. We were surprised to learn that more than 80% of the firm’s revenues come from outside Russia in a variety of different use cases as seen below.
NtechLab’s flagship technology – FindFace – is a powerful facial recognition technology with some impressive capabilities. It takes less than a second to search through a database as large as 1.5 billion images. And it’s not just mugshots. FindFace excels at identifying faces where not all information is made available. For example, their latest FindFace Security algorithms can identify a person when only 40% of their face is showing against a database of several hundred million images – in just 0.3 seconds.
It’s no surprise that FindFace is now being used by one of the world’s leading security camera companies – Genetec – which has more than 13,000 customers. Given the sensitivity around the use of facial recognition technology, NtechLab couldn’t tell us much about what they’re getting up to outside of Russia in places like Brazil, India, Malaysia, Spain, Portugal, and even China. That’s okay, because there are plenty of interesting things they’re doing in Russia we can talk about. Some of the things they’re doing are more obvious – like using facial recognition to identify shoplifters. FindFace is already being used across all major Russian cities to thwart shoplifters, but using facial recognition to catch shoplifters is something we’ve seen before. What we’re going to talk about today is way more interesting than that.
Facial Recognition for Events
Most Americans may not know what the FIFA World Cup is but it’s this big soccer game that takes place which makes the Super Bowl look like a Sunday afternoon pick-up game at the park. While the last men’s World Cup was being held in Moscow in 2018, someone walked off with the Budweiser sponsorship cup. Fortunately, they captured a photo of the thief and FindFace’s algorithms went to work scanning the millions of people attending the matches each day. The thief was actually dumb enough to come to a subsequent match and was detained by police. That use case demonstrates just how effective this technology can be to combat crime.
While people in the U.S. obsess over police body cams, Moscow police are moving past that in their technical prowess. NtechLab is currently working with the Moscow Department of Information Technology to develop a wearable AR solution for police officers. Currently, they’re using AR glasses from Epson that use FindFace facial recognition tech to actively scan faces and compare them against a database of individuals who are wanted by the police.
The hardware for the real-life pilot may change, but whatever they decide upon is expected to be rolled out to at least some police officers before the end of next year. And it’s also being deployed to fixed cameras as well.
Facial Recognition for Public Safety
In 2017, the city of Moscow upgraded its public surveillance system by adding NtechLab facial recognition technology to 1,500 CCTV surveillance cameras installed across the city. It’s only a pilot considering there are more than 160,000 CCTV cameras located across Moscow, but it’s already led to more than 200 arrests. This year, the mayor of Moscow met with President Putin to announce that the cameras that are placed near the entranceway of every living space in Moscow will be enabled with facial recognition algorithms such that they can be used to identify the activities of known criminals. There’s also a not-so-obvious benefit which involves the challenges of data archiving. By tracking faces and then storing the metadata only, a 10X to 100X efficiency can be realized by only storing relevant data. This means you can increase historical data for your camera feeds from 10 to 100 days without adding any storage.
It’s important to consider privacy implications which is probably front and center in most readers’ minds right now. “The Russians are watching my every move,” you’re probably thinking, but here’s something to consider. Firstly, the cameras have always been there. It’s always been possible to track people as long as the cameras have been in place. Facial recognition reduces the time it takes for the police to solve a crime. Time is the most important factor in solving crimes, and there’s a correlation between the time it takes to solve a crime and the likelihood a crime would be solved. Most cities around the world will not be forthright about their use of facial recognition algorithms, so assume that all cameras have such technology and behave accordingly. Sure, it can be used maliciously, but so can all technologies. Don’t buy it? Okay, then ban it. So, what happens to FindFace when facial recognition gets banned?
Beyond Facial Recognition
In the finance world, regulatory risk refers to situations where a firm can be impacted by changes in law that cannot be controlled. For example, facial recognition is already banned in several U.S. states like California and Massachusetts. Addressing this risk involves diversifying into applications beyond facial recognition. In order to understand the company’s strategy for mitigating regulatory risk, we spoke with Mr. Minin about other products under development in the company’s pipeline.
One area NtechLab is working on involves action detection such as identifying someone who is talking on a phone or moving a heavy object. (NtechLab finished top-3 in the NIST and IARPA challenge for detecting actions in video.) There are several interesting directions that such a technology can take. Sure, you might be able to detect people throwing punches or breaking windows, but security is far from the biggest opportunity here. Mr. Minin talked about something firms are interested in called “activity-based tracking.” For example, tracking the actions of a person in a manufacturing job that requires a series of repetitive steps can be used for quality assurance. (Remember our article on How to Detect Manufacturing Defects Using AI?) Then, there are dangerous industries like chemicals or nuclear energy where compliance people can be pacified by algorithms that snitch on people who don’t follow a process because they’ve done it a million times and think they know better. It’s all about tracking specific actions at specific times.
Other companies are working on activity-based tracking using computer vision, but Mr. Minin describes the challenges they’re running into as “low quality of recognition” and “very slow speeds.” NtechLab’s algorithms are extremely fast and scalable. In this year’s release, there’s literally no limit on the number of cameras that can be used at a single time. One of their major use cases at the moment is in the area of electricity production, but they’re also looking at more obscure use cases like healthcare. In countries where the mortality rate in hospitals is very high, algorithms can monitor hospital procedures and flag problems. Of course, the dirty little secret is that “activity-based tracking” is going to be used to train robots so this all becomes a non-issue eventually.
When you can’t look at the face anymore, you can simply just look at the body. It’s called silhouette detection, and after working on it for several years, NtechLab is getting to a final stage where the technology can be used in real-world scenarios. Silhouette detection is about examining your physical presence – the way you walk (also called your gait), your height, your massive beer gut, pretty much any attribute of your body that the cameras can capture.
When you combine all these attributes together, it’s pretty easy to track someone even if they try and change clothes or change their appearance. (We can tell when they’re changing clothes – remember activity based tracking?) People cannot fly or disappear. Even where there are no cameras, we can still pick them back up later when they cross the next available camera. If they step into a vehicle, we can track that too. Even if you’re one of the many people in Moscow who love to cover their plates while parked illegally, the camera can still capture the silhouette of the car. It’s just another factor that can be used to track you since cars can’t fly either – yet.
Famed Russian author Aleksandr Solzhenitsyn once said that, “For us in Russia, communism is a dead dog, while, for many people in the West, it is still a living lion.” He correctly points out that there’s a bad taste in the mouths of the Western world when it comes to their perceptions of Russia. It’s kind of like the bad taste we get in our mouths when listening to Americans bicker about politics. Leaving all that aside, there is a great deal of talent coming out of Russia that flies under the radar.
China’s leading facial recognition firm – Sensetime – has taken in $2.6 billion in funding. NtechLab is going head-to-head with Sensetime with just 0.0577% of that funding. It was a common theme we saw during our time in Russia – exciting startups with extremely small amounts of funding having a global impact. Maybe it’s time for venture capitalists around the world to take more notice of Mother Russia and her talented people. It was also Solzhenitsyn who once said, “Talent is always conscious of its own abundance, and does not object to sharing.”
We hope you enjoyed reading this article as much as we enjoyed writing it. Did you know you can hire our team of MBAs to write about your exciting disruptive tech story? Check it out.