Chinese based startup SenseTime uses AI to detect faces in a variety of ways for security and entertainment purposes.
In their most recent round of fundraising, the Chinese startup SenseTime raised $600 million. The recent funding approximately doubled the worth of this AI startup and their work is fascinating, or terrifying depending on who you ask.
Truthfully, SenseTime isn't much different than the startups I've covered here before. They offer facial recognition software, the same kind that tracks faces to tag you better in photographs or that allow surveillance cameras to detect and identify you to fine you for breaking litter or jaywalking rules. It's also the kind of startup that many would find creepy, especially after their most recent collaboration with the Chinese federal government.
I'm not here for doomsday prognoses, however, I'd really like to look at what SenseTime does so well that it's got the likes of Honda and Qualcomm interested in its technology.
SenseTime's tech is mostly focused on facial recognition software for various uses which we'll talk about a bit here. The main uses for their tech are static and dynamic facial recognition (using images). For companies, they also specialize in identification verification (using facial recognition), mobile internet and some features for autonomous driving vehicles.
I Always Feel Like Somebody's Watching Me
Let's take a look at their product offerings. Their SensePortrait-S Static Face Recognition Server is a facial recognition software which can detect faces from images and then extract, analyze and compare faces from photographs to those in a facial image database. This could be helpful when trying to tag a friend in a photo or identify a person from surveillance footage. The proposed uses of this technology on their website include verifying the identity of crime suspects, ID checks at public locations and security checks at various gateways like airports.
Their SensePortrait D Dynamic Face Recognition Server is similar to the product previously mentioned but uses facial recognition software for surveillance video streams, face tracking and facial comparison. It is best suited for security and criminal investigation of surveillance footage, visitor recognition and can even be used to mark the students in your 9 am lecture as present without having to do roll call.
Companies interested in using SenseTime's software have access to their Surveillance Platform which is powered by their deep learning algorithm. Using the Platform, users can use their video analysis tools to extract all relevant data, including data for population management. This can be used in addition to the Face Image Investigation System, which houses databases storing hundreds of millions of face images.
The Video Structuralisation Server is an intelligent engine for users to analyze footage. Users can even detect attributes like the gender, age, and clothing of people captured on video or vehicle license plates and car features (make, model etc.) just from a video.
Their mobile services are a little more light-hearted. Their Augmented Reality Rendering Engine and Platform tracks users' faces for entertainment purposes like live broadcasting and "image beautification" (like that super weird dog filter people seem to love). For protecting their devices, users also have access to SenseKeeper. Put simply, it's similar to how the new iPhones confront users with the ability to secure their device and even pay using their faces as "keys."
Although their main uses are surveillance, they also use computer vision technology for other business uses. For instance, SenseTime's SenseGo can be used in malls, supermarkets or even cafes to provide business owners with big data on customer flow and customer demographics. Their SenseDrive systems also allow for in-vehicle minoring which can detect when users are fatigued, distracted or when lane and road conditions are unsafe. Some of their SenseDrive systems can even send a warning signal when there may be a collision with an object, vehicle or pedestrian or when a particular path is dangerous.
The Good, The Bad, The Ugly
Reviewing the product offerings, it's important to look at the positives and negatives. For one, better, more accurate identification can mean better security and ease of identification. When investigating crimes or letting people through an open border it's important to know who everyone is as accurately as possible. Similarly, your time is too valuable for false identification and long verification times to slow you down when being identified.
Noticeably, however, their products are primarily used for surveillance and crime investigation. The use of technology by the government and law enforcement forces employed by the government for the purpose of tracking, detecting and punishing the human populace is a concern for many, especially as the level of surveillance around the world is becoming more frightening and conspicuous.
In the age of governments spying on their citizens and large social media platforms selling and trading sensitive information, security has become less of an expectation and more of an ideal. The concerns of many that perhaps the most surveilled populace on the planet can become more tracked leaves consumers feeling like rats in a maze more than potential customers.
I can't begin to promise an end to a reality like this or unequivocally support or bash SenseTime for their product development. Facial recognition software can seem creepy and weird but it also keeps us safe and saves us time by making identification processes simpler and more accurate. The best way to keep governments and companies in check may be to vote in more aggressive policies for information handling and take more decisive action against those who breach public trust but for now, the feeling of being watched isn't going away any time soon.