Facial Recognition in China with SenseTime
If you’ve spent any meaningful amount of time in a managerial role, you probably understand the importance of having a team with members who feel comfortable speaking up when they disagree with you. That’s a critical component of high-performing teams, and one you take for granted when you start to manage teams of Chinese people. In addition to everyone being agreeable, there are other things that Western people will find difficult about managing teams of Chinese. It’s not the long lunches, it’s not that they like to sleep at their desks, and it’s not all the cute cartoon figurines they place on their computer monitors. It’s the fact that initially, it’s hard to tell them apart.
Yes, we said it. Put most Western people in a room full of Chinese people and they’ll have a hard time telling them apart – initially. If you mention that to the Chinese, they’ll nod thoughtfully in agreement and tell you that to them, Western people all look the same too – like that character on Friends. This is because our brains were “trained” to recognize facial differences using different data sets. That’s also why when it comes to using artificial intelligence to identify facial differences, one size algorithm doesn’t fit all.
When it comes to applying facial recognition in China, the country seems to be farther ahead than any other. Just look at some of these examples:
- The Shanghai metro is developing facial recognition systems that will be placed at the entrance of each subway to verify the identity of commuters
- A new police car can now do a 360-degree scan to identify faces at up to 60 yards away while traveling at 75 miles per hour
- Railway police use facial-recognition eyeware to identify someone in just 100 milliseconds from a database of 10,000 individuals
- Unmanned convenience stores use facial recognition for payments, while KFC uses “smile-to-pay” technology
- China’s answer to Airbnb, Xiaozhu, will soon use facial recognition for check-ins
- Chinese exam authorities are using facial recognition to catch cheaters for competitive college entrance exams
And the list goes on. Meanwhile in the USA, everyone is up in arms over the fact that the algorithms can’t consistently identify people who have different skin colors. With few barriers to slow things down, and a lot of hard 9-9-6 work, the facial recognition market in China is rapidly reaching maturity. One company that’s managed to achieve unicorn status by telling Chinese people apart is SenseTime.
Founded in 2014, Hong Kong startup SenseTime has taken in a whopping $637 million in funding from investors that include large corporates like Alibaba and Qualcomm. Their latest funding round of $227 million closed in November of last year and cemented SenseTime’s status as a unicorn moving the company from a $1.47 billion valuation to a $3 billion valuation according to China Money Network. That would imply that the company’s valuation doubled in just 4 months which meshes well with the fact that China is now taking in more AI funding globally than any other country:
The extent to which China plans to use AI is evident in the variety of use cases that SenseTime alone offers up. They aren’t just doing face recognition but also object recognition, vehicle recognition, and image processing. Let’s take a look at some examples of SenseTime’s technology being put to good use.
Catching Bad Guys
With a population nearly 3X the size of Los Angeles, it’s safe to say that Guangzhou has a few criminals running around doing bad things. In 2017, the police established a “video detection department” which uses SenseTotem to compare crime scene facial captures with database photos.
The system has since identified more than 2,000 suspects, and in turn, managed to capture more than 800 people with nearly 100 criminal cases being successfully solved. It’s a good reason to think twice before trying to steal designer sunglasses.
SensePhoto is a smartphone image processing solution that helps capture better photos which could be the deciding factor for avid smartphone photographers everywhere. QiKU is a Chinese smartphone with its own Android operating system that has a “binocular solution” which addresses the zoom problem that many smartphones have.
Other features include things like automatic face grouping to make organizing your albums easier. Weibo, China’s largest recreational social platform with 340 million active monthly users, uses SenseTime technology for functions like face effects, beautification, and gesture recognition.
You may have never heard of Rong360, but they’re another Chinese unicorn that actually went public in the USA late last year under the name Jianpu Technology Inc (NYSE:JT). With the large influx of Chinese moving into the “middle class”, Rong360 has become the leading independent open platform for discovery and recommendation of financial products in China.
SenseTime provides their SenseID identity verification service to Rong360 so that they can offer “fast loan products” which probably have corresponding interest rates that accrue interest just as fast as the loans are being given out. The product also offers “liveliness detection” to make sure you’re not just using a static photo to trick the verification algorithms.
Spread across more than 600 cities, China’s largest retail enterprise, Suning, is now working with SenseTime to explore the application of face identification for unmanned shopping (remember Bodega?), membership management, and payment verification, along with the use of artificial intelligence to create personalized shopping experiences.
The Chinese are less likely to become alarmed when they walk into their local store and a robot standing at the entryway greets them with a “Hello Mr. Wong, do you need some movies to go along with that DVD player your wife ordered online yesterday?”.
Better Weather Forecasts
We’ve talked before about how useful big data from satellites can be in doing what we refer to as geospatial analysis. It’s not just for looking to see where the deforestation problems are occurring, or finding out how much oil China is storing, but also for more tactical applications – like the weather. The China Meteorological Administration is using SenseRemote, an intelligent remote sensing image solution developed by SenseTime, to improve the weather data coming from satellites to allow for more accurate forecasting.
In addition to that, SenseTime is also working on information extraction from satellite images to look at things like road conditions, floodwaters, snow accumulation, and other meteorological and environmental elements worth analyzing.
In addition to the use cases mentioned above, there are a plethora of other applications that SenseTime addresses with their AI platform. Their partnership with Honda that was announced late last year is a “long-term development agreement” which will focus on building “smart AI cars with autonomous driving” that couple SenseTime’s imaging capabilities with Honda’s voice recognition capabilities.
One might wonder what SenseTime is doing with all that money they’ve raised and the answer is that they are investing in infrastructure and intellectual property. They’ve established China’s largest and only self-developed deep learning supercomputing center, and applied for over 500 patents, 90% of which are original inventions. With over 400 clients using their technology, they’ve entered into key partnerships with large technology companies like Qualcomm, Nvidia, and smartphone company Xiaomi.
SenseTime is not alone though in their quest to dominate facial recognition in China. An article by MIT Technology Review mentions how “nearly 90 percent of China’s roughly 200 top Internet companies use Face ID” and that the “world’s largest face-recognition technology platform, currently used by more than 300,000 developers in 150 countries” is actually another Chinese unicorn called Megvi with their facial recognition platform called Face++. Given that there are nearly 1.4 billion people in China, it’s safe to say that the market is large enough for multiple players. One thing seems for sure though. It’s not likely to be a market where many American companies will be operating in.