Trials in London use facial recognition to get you your beer faster at the bar, but then also track you as you go home on the train.
Facial recognition is getting lots of attention here in the UK, as we’re seeing everything from the wacky to the scary. The wacky involves a bar equipped with facial recognition technology to help Brits order an estimated 78 million more pints of beer per year, and the scary includes train commuters and workers being tracked using facial recognition at a major London train station without them knowing it.
And it’s going to get even easier to install facial recognition technology, with Amazon this week announcing the addition of ‘fear’ to its emotion detection through its Rekognition software. With its latest release, Amazon said it has improved the accuracy of gender identification; and improved accuracy for emotion detection of seven emotions: ‘happy’, ‘sad’, ‘angry’, ‘surprised’, ‘disgusted’, ‘calm’ and ‘confused’, as well as adding a new emotion, ‘fear’. In this release it has also improved age range estimation accuracy.
Amazon says no machine learning experience is needed, which in theory means anyone could add facial recognition analysis to their system. All they need to do is send images or video to its API, and Amazon can identify objects and people and provide facial recognition and analysis. What’s more it learns from all the data being fed into its system, so it gets even better datasets for learning.
Use my face to order a pint — fast
So, no wonder we get all kinds of applications being deployed. Like the bar in London which has just trialed facial recognition to place drinkers at a busy bar in a dynamically intelligent queue. Yes, as the press release from the firm that developed the system, DataSparq, said, the new tech will eliminate queue jumping in bars and pubs. It said the biggest complaint for British drinkers is people pushing in front of them while they are waiting at the bar to order a drink, and that more than three quarters of Brits have walked out of busy bars due to long waiting times.
In the survey of 2,000 British drinkers commissioned by DataSparq, it said that nearly two-thirds would stay in the pub and order if they knew exactly how long the wait to be served was. That’s why the company developed the artificial intelligence (AI) bar, which it said gives customers this clarity.
The AI bar, trialed at a bar in London (Image: DataSparq)
So, what exactly does the AI bar do with the facial recognition? Well, in a crowded bar, it will make ordering faster and fairer by identifying faces and then putting them in an ‘intelligent’ virtual queue. It will also carry out faster ID checks — if anyone looks under-age, the system will prompt them to produce ID before they reach the front of the queue.
Additional functionality currently being developed includes the ability for customers to re-order their drinks while still in the queue. The system memorizes drinkers’ orders and reacts to simple hand signals if there are any changes. Another feature will include a ‘FaceTab’ — a mechanism that visually adds people to a bar tab, only allowing certain faces to order against that tab.
The AI bar will be offered as a software-as-a-service (SaaS) product and works by simply using a standard webcam, display screen and internet connection.
Clearly, this triggered our curiosity (though we didn’t get to the trial at the bar), so we spoke to DataSparq’s managing director, John Wyllie, who also happens to be a computer scientist from the University of Cambridge, to find out more about the AI bar. He told us the trial at the London bar was simply a technical rehearsal — the real thing will be launching probably next month, and the company is already talking to drinks companies and pub owners to roll out the technology nationwide over the next 12 months. He said, “The system can be installed anywhere and is scalable — so we are expecting it to start in bars and progress into music festivals and beyond.”
He told us there’s a lot of AI hype, but all it does is just use databases to solve specific problems. In that respect he said, “We’ve tried to use as much commoditized tech as possible, and piece it together.” They had started developing the solution in May 2019 using “a whole bunch of tools which are readily available for face detection, recognition and analysis”. He said they debated what data should be sent to the cloud and what should be done locally. “We do a lot locally.” The reason is that the response time needs to be quick to respond to customer demand at the bar. The face recognition API is quick, but the face comparison is more time consuming so that is sent to the cloud.
The company uses a simple black box based on a Raspberry Pi, which connects to a local server, taking the input from a standard webcam and with just three interfaces — a USB input, Ethernet connection, and a HDMI output. On the topic of protecting user’s identity, Wyllie stresses that they are not storing data after the bar closes at night, and all it’s doing is storing faces and comparing them for that particular day.
Tracking customers and workers at King’s Cross Station
Where facial recognition has created controversy, though, is the revelation over the last week that one of London’s busiest railway stations, King’s Cross, has been using facial recognition to track visitors in the name of ensuring public safety. This is no small issue — the 67-acre site owned by King’s Cross includes London-based development centers for Facebook and Google, and Samsung also has facilities there. The property developer of the site, Argent, said in its own documentation that 176 million people pass through the station annually.
According to civil liberties group Big Brother Watch, it managed to identify a camera at the station which features advanced pattern detection and self-learning video analytics embedded in the camera, with up to 16 MP resolution. Speaking to the Financial Times, Stephanie Hare, a researcher of facial recognition technologies, said that it’s a worrying development for anyone who doesn’t want to participate, since they don’t have a choice of opting out. She commented, “You can’t opt out of walking around London, or working there. How do they defend it when this technology is the subject of legal action and MPs are calling for a moratorium on it?”
Already, earlier in the year there was an episode in London where a man was stopped and fined £90 because he covered his face to avoid being scanned by a facial recognition trial being carried out in a public place by the local police force. According to civil rights campaigners, there was nothing suspicious about him, he was just exercising his right not to have his face scanned by the cameras.
In London and the UK, we’ve grown accustomed to surveillance throughout our public places for many years now — they were a key tool in identifying the movements and identities of the suicide bombers involved in the London underground bombings on 7 July 2005, and in many other high profile police cases since.
This means cameras are everywhere, and the British public accept it as a part of daily life. But with feature creep, and with facial recognition technologies getting so accurate, will we still be so accepting when we know that Big Brother is not just always watching you, but can also know your emotions and how you feel, and possibly even fall victim to mistaken identity?
We might be overjoyed at getting that pint of beer faster with the AI bar, but then as you get on the train at the station, ‘they’ will know if you’ve had too many.