Access control specialist Brivo has launched a facial detection tool, significantly speeding up the process of identifying people going in and out of entry points.
Brivo Snapshot evaluates every frame of security video, scoring them based on the amount of face that it detects and picking out the most relevant ones. The algorithm allows the software to create a photo album that records “the complete event-related visual setting” – a record of everyone that passed an entry point.
This means that every situation of potential interest around entry points is already pre-analysed, pre-indexed and immediately available when someone wants to see what happened.
In the past, someone would have had to spend 15-40 seconds scrolling through a video to find one frame of information because access control and video systems were developed separately and were not always in sync, Steve Van Till, CEO and founder of Brivo, told PlaceTech.
“If you’re doing this three times a day, you don’t really care that it takes you a minute every time,” he said. “If you’re doing this hundreds of times per day, those minutes add up and they become an almost impossible burden for getting value out of integrating video and access control.”
What about facial recognition?
At the moment, Snapshot simply detects faces, rather than recognising them. But Van Till said it would be a “fairly short step” to build in recognition and add names to faces.
With that capability, in theory, Brivo would be able to know if it has seen a unique face or a regular employee. That would allow the system to recognise specific people and let them in without them having to present an access card.
Whether or not Brivo takes that step is not clear. Van Till said: “Is that a feature that customers want? I’m not sure. That would be one of the next things we find out.”
Taking security automation further
Having automated facial detection, Brivo will soon release an anomaly detection algorithm, which will be able to learn what is normal in a specific building and notify the facility’s team if something abnormal happens. That might be someone entering the building at midnight or approaching an entry point in a group for the first time in five years – anything that, in context, seems unusual.
The first step would be a ‘data ingestion period’ where the system gets to know the building – or taps into historical data if that is available. Users would then set a threshold for abnormal events that controls the point at which Brivo notifies the team if something unusual happens.
Van Till said: “What we’re excited about here is that we do have the ability now to have what used to be called intelligent alarm filtering, but it’s based on the particulars of your situation, as opposed to hand-built rules, which is how most of those engines worked in the past.”
While tools like Snapshot are built on AI and machine learning, Van Till said that the function is more important than the fact that it’s AI: “We regard [machine learning] as a tool like any other tool. We’ve got algorithms galore that are or are not machine learning. Every one has a purpose, that purpose needs to be tied to making life better for the users. And that’s what we’ve done again here.”