The Center for Data Innovation spoke with Max Constant, chief commercial officer and spokesperson of AnyVision, a company based in Israel, which uses AI to provide real-time face, object, and human recognition in mass crowd events and public areas. Constant discussed the diverse applications of face recognition technology and the challenges it can address in various areas.
Eline Chivot: What are some of the problems AnyVision is trying to solve?
Max Constant: The main problem that AnyVision is trying to solve is to make AI accessible to the world by enabling all cameras to be smart. In doing so, we focus on three main disciplines within computer vision: faces, bodies, and objects.
Likewise, we’re focused on making our solution truly plug-and-play by being agnostic to both the cameras in which those analytics operate and the computational framework required to actually function in the environment and use case in which it’s expected to operate.
Chivot: How have improvements in cameras impacted human and objection recognition?
Constant: Like any other space, technology is always improving. But there are two distinct markets within the camera space: Existing cameras that have been in the market for 10 or 20 years, and smart cameras of the future that will be coming online in the next 1 to 10 years. We understand that both markets are equally compelling and have constructed our design framework to operate on both types of sensors, regardless of whether they are simple storage and recording devices, or if they have more exquisite capabilities in the future. Part of making AI accessible to the world and making every camera smart is to truly be able to function on whatever camera, whatever compute, or whatever environment is available to the customer.
Chivot: How has facial recognition technology improved over the past few years? What capabilities are available now that were not available a decade ago?
Constant: Aside from the normal metrics used to benchmark facial recognition (true positive, false positive, etc.), what we’ve seen in previous years that has been really pioneered by AnyVision’s CTO and our research team is a fundamental rethinking of how computer vision-related problems should be approached. By solving those underlying problems within computer vision, AnyVision has been able to unlock a world that previously did not exist. In time, that will lead to better accuracy rates but also better, more efficient computation, and significantly more use cases because we will have actually solved the underlying problems in computer vision, based on the environments and use cases to which they are applicable.
Chivot: The applications of your technology are surprisingly diverse including airports, stadiums, shopping malls, and casinos. What are some examples which show how using AnyVision can be particularly valuable for security and commercial purposes?
Constant: We are a core technology company that builds products to solve a variety of use cases across different verticals. Fundamental to that thesis is: If you can crack the face, body, or object problem in a thoughtful way, those same faces, bodies, or objects can be used and leveraged to solve a diverse set of use cases beyond just security and surveillance.
For us, security and surveillance are inclusive of both the “white list” and “restricted list” use cases. The “restricted list” may include shoplifters in stores, card counters in casinos, and wanted individuals on other watch lists. That same platform, however, can be leveraged to do things such as fast entry into stadiums, employee and vendor tracking to ensure compliance, etc. The idea here is that if you truly build a core technology upon which multiple products can be built, to solve unique use cases across different verticals, you can start to become a multi-use technology and multi-use company overnight.
Chivot: How does facial recognition compare to other biometrics (in terms of accuracy and usability)? And what are the main technical challenges your company hopes to solve in the coming years?
Constant: There is no fool-proof, 100 percent biometric solution. However, given those constraints, we do believe that the biometrics world will evolve in a way that will require dual-authentication in order to have a truly comprehensive solution. We are pushing the boundaries on how the world is thinking about authentication using the face, given the fact that we have built a solution that not only has a very high true positive and a very low false negative rate, but it’s also flexible across the environments that individuals are expected to travel through. Meaning, why shouldn’t your face be the key to your car? Why shouldn’t your face be the key to your bank account? Why shouldn’t your face be the key to your office? In short, your identity should be your key, regardless of where you are and what you’re trying to do, assuming that you have the correct true positive rates and almost no false negative rates.
Outside of continuing to push the boundaries of what is possible from an analytics standpoint, of equal importance is pushing the boundaries of what was thought possible from a computational and sensor standpoint. In doing so, we can continue to open up new fields or disciplines of research and applicability of this technology, beyond just security and surveillance, retail analytics, or authentication. We can move into areas like medicine, sports, automotive, engineering, etc. Ultimately, we’re looking to understand how vision and signal processing is going to impact any sort of IoT-connected camera in the future, as well as the products that will be able to solve everyday problems and make the world a better place.