Who’s recognizing faces in China and Russia and why they’re doing it
Let’s call a spade a spade
When it comes to Facial recognition systems, it’s helpful to think of a spade as a metaphor — you can use it to clear snow from a road, path or cycleway and to hack your neighbor to death. In either case, the spade itself is just a tool. Face recognition is just a tool as well. One can use it both to the good and bad.
The tool that’s not spotlessly efficient. Don’t forget, that detection accuracy depends on a variety of factors, from the weather to lighting.
First and foremost, Facial recognition technology is the part of Video analytics, solution based on AI, Big Data and Cloud Video Surveillance for businesses, from banks and retail chains to coffee shops and small beauty salons.
Companies still install Video Surveillance systems more for their own security than because Big Brother told them to.
Security cameras shoot about 83 million Full HD films every month. All of them represent data for analysis.
Yesterday’s “dumb” cameras are being replaced by smart ones. The software for video analytics can be pre-installed on the camera by the manufacturer. The alternative is video analytics in the cloud, which can work even with a relatively cheap camera. The second option may be better for the user since cameras with the software for video analytics “on board” are much more expensive.
Cameras have moved beyond functioning as “eyes” and become smart AI-analysts, sending push notifications when necessary. They let the manager know when there is a queue in a store or somebody is robbing a cash register; whether employees are working too hard, not wearing hard hats, or are they always late. Even beyond that, the system can use collected data to predict problems and recognize thieves that came back for a second round. Video analytics can also help to monitor the availability of popular goods on store shelves.
So, what we ultimately have is a powerful and enormous global Video analytics market where the best minds, money, manufacturing resources, and jobs are concentrated, one that in 2016 was estimated to be worth a cool $1.6 billion. And the forecast is for it to grow to $8.55 billion by 2023 and $9.4 billion by 2025.
According to MarketsandMarkets data, the fastest growth rates of the Video analytics market will be in Russia and China.
Picture caption: Recognized face, Elena Ivanova, Female, 25 years old, Detection Accuracy 98%
Technology as the highest form of imperialism
The Chinese can get the spotlight of shame turned on them through no fault of their own as happened with one well-known businesswoman. A camera mistakenly identified a picture of her face on the side of a bus driving by as the original and submitted it to the “shame screen” with a notification that she’d crossed at a red light.
There is also a chance that social activists or women openly supporting #MeToo movement, which is frowned upon in China, won’t be able to hide from Big Brother as well.
Building a Video Surveillance system on par with what the Chinese have, is practically impossible. It was back in 2005 that they started implementing state Video Surveillance, calling it Skynet either because of an overenthusiastic sense of self-irony or a complete lack of it.
The ten million people in the average Chinese city are watched by between several hundreds of thousands and a million security cameras. At the end of 2017, there were more than 170 million cameras in the country, with around 400 million more planned for the following three years. And the Chinese certainly know how to follow through on their plans.
By 2020, the target is to have complete Video Surveillance of key public areas in a bid to eliminate blind spots. It’s almost like the cameras were set up in “several layers”.
Don’t forget that China is the world’s largest manufacturer of hardware for Video Surveillance. It’s the home country of giants like Dahua and Hikvision. Both companies get state contracts, and more than half of the cameras they produce are earmarked for the domestic market.
Expenditures on security, which includes Video Surveillance, for China’s largest autonomous region of Xinjiang jumped from 30.05 billion Yuan in 2016 to 57.95 billion Yuan in 2017.
Hikvision sold 55 million cameras domestically in 2016, that number growing to 70 million in 2017 and 90 million in 2018. Given that more than 70% of Hikvision’s sales are in China, the company is projected to sell 250 million cameras there between 2017 and 2020.
In other words, there will be one new camera from that manufacturer alone for every five Chinese citizens. Let’s put the cherry on top: Hikvision only controls half the Chinese market.
Xu Hugh, a Dahua representative, notes that “facial recognition capabilities in China currently surpass those of the human eye,” attributing that to, among other things, the Chinese switch from 2D Face Recognition to 3D Face Recognition. In the first case, algorithms are used to analyze 2D images stored in a database. 3D recognition analyzes reconstructed 3D images and has proven to be much more accurate.
Chinese police forces use smart glasses with cameras and Facial recognition to catch criminals. However, some data, including the scale of the criminal databases the glasses work with as well as the number of false detections, is generally not disclosed to media.
You can scan your face in China to make purchases (pay for food at KFC, for example), make payments, and enter buildings. With Alipay one only needs to smile so the payment system knows there is a real person, not just a picture. It’s been shown to be impossible to fool Alipay by changing your hair color, putting on makeup, or using a wig. The system uses a set of distinctive attributes that take into account facial geometry and set points distributed around it.
The multistep scanning process takes between one and two seconds. In that time, the 3D camera runs an algorithm for detecting living images, executes a scan, and analyzes shadows and other attributes to make sure a video isn’t being used in an attempt to fool the system.
The video analytics are complemented by audio analytics. The tech gives access to social services, lets one check in at hotels, and is also used to buy train tickets remotely.
The province of Henan uses Facial recognition at the entrance to the Longmen Grottoes. Visitors have to download the app provided by the Longmen Grottoes, buy an e-ticket, take a picture of their face, and enter their personal information. The result is never having to wait in line.
The number of cameras on Chinese streets outpaces anything Russian authorities have done so far. They record everything, from highways to sidewalks. And if we take a look at roads and especially intersections, there’s an unbelievable number of cameras working in conjunction with the smart city system to monitor speeds, check blacklists for people and cars, identify parking violations, and even flag illegal street traders.
By way of comparison, Moscow authorities, on January 11, 2019, suggestedthat criminals give the city a wide berth because there was nowhere to hide thanks to the 167,000 cameras of the Moscow Video Surveillance system.
Moscow Police Chief Oleg Baranov claims security cameras helped to catch 3,000 criminals in 2018. The federal “Safe Region” system has about 23,000 cameras hooked up to it. And finally, there are 17,000 cameras in the Moscow subway, probably the location most discussed in the media.
The Government of Moscow announced that they had used the existing system to monitor more than 17,500 public events and identify problems with public cleanup in residential areas. News about pilot programs and experiments with Facial recognition appear regularly, from a school in Yaroslavl to Moscow Domodedovo Airport and buses in St. Petersburg. But those are still more the exception than the rule.
Why it’s Russian business, and not the public sector, that could “catch” China
The numbers above show the stark difference in investment in Russia and China. Still, providers going beyond public contracts to develop Video analytics solutions for business are enjoying major capital influxes, including those from government funds.
It couldn’t be more logical: some of those companies and startups do end up making a return on investments. While it wasn’t that long ago that only Large Enterprise could afford Facial recognition, in 2019 Video analytics has become much more available for SMB and SME.
China is recognized as a leader when it comes to manufacturing hardware for Video Surveillance. Russia, by contrast, has proven to be the source of intellectual resources such as programmers and engineers.
As they work with Chinese manufacturers and develop camera software, Russian companies have the opportunity to make good money and secure a strong market position. The example is Ivideon, the web service that provides both cloud Video Surveillance and Video analytics solutions. In 2019 Ivideon signed a deal as a strategic partner for the Chinese giant Dahua Technology.
When it comes to Facial recognition as a product, each segment in business has its own needs, both in quantity and price. The larger the company, the more cameras they may need. Solutions for Large Enterprise are always tricky and customized, with that customization coming at a cost. Small and mid-sized companies, on the other hand, can get away with one camera with Facial recognition. And that’s the way the market stands right now: a single cloud service is fully capable of handling clients varying in their scale, pricing, and customization level.
This trend is quite universal and applicable not only to Russia and China but also to Europe and the US. Ivideon is already testing its own Face recognition technology in Russia, soon it will be available for business all around the world.