Traditionally, facial recognition solutions have had two primary applications in the security industry: being leveraged as a biometric identifier in high-end access control deployments and aiding in real-time video surveillance investigations. The fact is, however, we live in a time when security systems must deliver a greater return-on-investment than the primary functions for which they were created. The use cases for video surveillance, for example, have expanded well-beyond post-security event analysis and are now being leveraged by a variety of end-users to improve business operations.
It is these outside-the-box applications that are driving much of the excitement around some of the emerging technologies in the industry, including facial recognition. While privacy advocates have done their best to quash the use of these systems by government agencies, developers have been undeterred and the technology continues to advance at a rapid pace leading to a whole host of new market entrants in recent years.
Among these new players include CyberLink, a Taiwanese-based company that is making its debut appearance at GSX (Booth # 2394) this year with its AI-powered facial recognition engine called “FaceMe.” Although they may not be a household name in security, Richard Carriere, CyberLink’s SVP of Global Marketing, says that the company itself has actually been around for over two decades.
“If you’ve used a Windows PC in your life, you’ve almost certainly used one of our solutions for either video playback or for photo/video editing. This is where we come from, so for years we’ve been bundling billions of copies of our digital media software on PCs, which evolved to mobile apps in the last four or five years,” Carriere explains.
It was this migration to mobile apps that led the company to develop its facial recognition product, which was initially not security-focused at all.
“Our mobile apps were related to photo/video editing of augmented reality and one of the very popular apps we have that’s now spun out in a subsidiary is an app called YouCam Makeup, which is a virtual makeup app that has been installed over 800 million times around the world. Most cosmetic companies now offer a digital selection of their products and the reason why they trust us is that we were able – using AI and facial recognition – to map extremely precisely the faces of people through vectors measured on the face,” he says. “A company like Chanel, for example, they want people to try their expensive lipstick virtually on an app and they want to make sure it will look natural.”
Facial Recognition on the Edge
Unlike many companies that have to run their facial recognition solution on a server, CyberLink, through its experience in developing the makeup app, was forced to make the technology run on the edge, which has given them the ability to perform facial recognition both quickly and accurately. In fact, the company’s FaceMe engine recently ranked 12th among all participants in a facial recognition accuracy test performed by the National Institute of Standards and Technology (NIST) with a 97.02% accuracy rate.
“I would say there are three things that differentiate us: the fact that we run on the edge, across platforms and on any device. No one is as flexible as we are with their solutions – period,” Carriere says. “The speed of running at the edge is a huge advantage. Another big advantage is our precision. You need to make sure when you implement solutions that you’re extremely good at identifying people and depending on how you set up our model – anywhere between 95% and 98.92% precision to identify a person – the risk of identifying the wrong person has the right one being one in a million, that is a big advantage.”
Applications for Security and Beyond
In a demonstration of the company’s FaceMe engine, Craig Campbell, Sales Engineer for CyberLink, showed how, even without a person’s ID on file, the solution can provide a wealth of information about an individual, including their age, gender, and even the emotions (happy, sad, mad, etc.) they’re expressing facially with an amazing degree of accuracy. According to Carriere, this could be used by a retailer to deliver targeted ads in a store as customers enter.
“Several things could be done with digital signage depending on who's looking at the sign,” he explains. “You could either just collect statistics or see how people react or you can dynamically personalize the signage. Let’s say I walk in Macy’s, I’m a middle-aged man and there are signs offering special deals; I might be offered something different than the teenage girl who walks in the store.”
The system could also be used to identify people on a blacklist, such as shoplifters, or to even lock down a store to prevent a potential robbery. “In retail environments where someone shows up wearing a hood over their head to hide their face, there’s a possibility to set a trigger and lock the door before people come in,” Carriere says.
For the first time at GSX, the company will debut its U Alerts solution, which works with FaceMe to notify users when a visitor or person of interest is captured on camera within a facility. “Let’s say you’re expecting a customer or visitor and you have their photo or digital ID and when they show up, you can be notified directly,” he says.
While there are obviously broad applications for facial recognition, Carriere says that security is getting the most traction right now, particularly for facility access, visitor notification, etc.
“Where that is evolving is in looking for specific people, let’s say VIPs or people who are not welcome… and it is also evolving the broad context of security for employees clocking in and clocking out because when you use someone’s face you don’t have instances where employees might cheat by clocking in their friends,” he adds. “Also access to restricted facilities, how much time people are spending in some areas that are restricted or at risk. In industrial environments, for example, people can only be exposed to certain substances for so much time.”
Though many people view facial recognition as a tool that can be leveraged to steal their identity, Carriere says that the opposite is actually true and that it will protect identities moving forward.
“People hear AI they think that robots are taking over. There’s no robot hidden here, it’s just data that helps connect the dots to understand when vectors are read on the face of someone to tell who they are,” he says. “When people get more comfortable and learn more about that, it should change the landscape.”
About the Author:
Joel Griffin is the Editor of SecurityInfoWatch.com and a veteran security journalist. You can reach him at [email protected].
