How AI Is Shaking Stadium Security Awake

The starting point for any event is making sure that it’s a safe environment for fans. New developments in facial recognition technology make it easier to improve the safety of venues and events, catch wrongdoers and weed out troublemakers who spoil the fun. But improved security can be just the start of a journey toward a smoother and more enjoyable match-day experience for fans and families. We spoke to Corsight AI President Rob Watts about how the landscape is changing.

Facial recognition is a highly specialized space. You have been involved in it for a long while.

Yes, and that’s why I’m at Corsight AI now. I was leading NEC’s business across Europe. We were doing facial recognition. I’d been involved in policing, intelligence, defence industry for about 20 years prior to that. When Corsight came to me about five years ago and asked me to become chief exec they told me their software could do things that I honestly did not believe were possible. But they were absolutely right. The Corsight technology approaches facial recognition in a way that’s completely different from everybody else.  

Can you explain it in a nutshell?

Almost without exception other systems started with put your passport on the slab and we’ll recognise you. Corsight started with trying to identify people throwing Molotov cocktails at the military with hoods on and scarves on. What other systems are doing is plotting dots on the face. This is what’s called feature-based facial recognition. It works like fingerprints for the face. They’ll plot points on that face to be able to match to a similar face. We don’t do that. What we do is in effect we build a mesh of data points across the face for everybody. We can see people at 90° head turn with face masks with sunglasses with hats on. We only need 50 pixels between the ears and half of the face, sometimes even less. Other systems fail when you’ve got a head turn greater than 30°. We can recognize people at 90° head turn. Others fail when you’ve got greater than 40° rake on a camera. We can handle a 70° angle.

So it’ s about a deeper capability for identifying people.

With more accuracy. And in real time. We’re not just talking about identifying people. We’re talking about patterns of behaviour too. Think of someone coming back to a stadium and going around five, six, nine, ten 10 entrances on the same day. That’s something you might want to be aware of. Why is that happening? You can have as many security guards as you want patrolling the stadium. They’re not going to pick that up. Corsight does it instantly. It’s not just about real-time security. If there is bad behaviour by fans, we can identify the people responsible. We can then track them back to their point of entry. We track that back to the ticket. We know who’s bought that ticket and we can pay them a visit in the morning. We were involved with a country which had been banned from competitions because of right-wing or racist behaviour by fans. They put our system in and it helped to get back into the FIFA ranks. 

So security is often about what the venue or the authorities do after the fact. 

They no longer have to go through the video from the cameras bit by bit to find someone or an incident. That’s archaic and hugely labour-intensive. Corsight does it 30 times faster and it’s more accurate. We’re not talking about replacing or installing new cameras. Corsight plugs the feeds from their existing cameras into our software to give them facial recognition in real time and post-event. And it’s important to understand that this isn’t just about security. Sure, that is the foundation. But think of it as intelligence. There’s an intelligence dividend for the venue.

What does that mean?

A good example is one of the big malls in the U.S.  They have hundreds of retailers as their tenants and they’re giving them not just security information, but also footfall information. What’s the age and gender of the people that have gone past their store? What’s the dwell time of people outside their store? Or, say you have teams of workers being supplied by contractors. We did some work recently where the company said they were sending ten carpenters and only two were actually showing up. Sports venues have hundreds of people working on site. We enable them to see who’s there and if they’re in the right place doing what they’re supposed to be doing. We’re a holistic solution. We tell all our customers that we give them three time return on their investment and the facts bear us out.

Why have sports venues been slow to adopt facial recognition technology?

Two main reasons. The first one is racial bias. The problem is that facial recognition wrongly identifies a black male more frequently than a white male. Corsight has put a huge amount of investment into the reduction of racial bias in our software, to the point where I state with confidence that the problem has been eliminated. That is a big differentiator in our favour. The other obstacle is privacy. Again, our software, in real time, blurs out the faces of people who are not relevant in any given circumstances. We are about identifying the right face not other faces. We safeguard privacy.

There must be use cases where a fan might want to be identified. Suppose a disabled person wants the venue to know that he’s arriving so he get some assistance?

Exactly. Or VIPs. Or just fans who want to be able to access an area or a service. We are just at the start. Facial recognition is a game-changer. You have a face, why do you need to stand in a security queue? You have a face, why do you need a ticket? You have a face, why do you need a credit card to pay for a beer at halftime? It’s true that the market has progressed by leaps and bounds. We now have QR codes on the phone and all of that sort of stuff. But it’s still archaic. We at Corsight are sitting on the sidelines saying, look, you can do all of this stuff if you want to, but realistically, let’s start with what security. There’s so much that needs to be done and we can help you now.