Manufacturer 1-on-1: SAFR’s Dan Grimm
Facial recognition solutions have experienced what some may say is a renaissance recently across the security industry. While the technology has been around for years, advancements in video analytics have made it much more accurate and have thus increased its adoption for a variety of different applications.
In many ways, the industry is only beginning to scratch the surface of what is possible with facial recognition, but it is already being touted as a potential game-changer for the market. From being able to spot criminal suspects in crowds of people to opening office doors, an ever-growing number of organizations are leveraging face recognition to bolster security and improve operations.
However, as with many biometric products, there has also been a growing chorus of critics who say the technology presents a grave privacy threat to users. With these concerns in mind, facial recognition providers are working to balance privacy and ethics with increasing market demand, which has surged exponentially in the wake of the COVID-19 pandemic.
This increased demand has also led to a number of new entrants looking to carve out their own niche in the marketplace. Among these companies is SAFR, a provider of AI-powered facial recognition for live video analysis.
SecurityInfoWatch.com (SIW) recently sat down with Dan Grimm, General Manager at SAFR, which is part of RealNetworks, Inc., to discuss what differentiates their technology from competitive offerings and some of the biggest issues currently facing the industry.
SIW: What was the genesis of SAFR and how was the technology originally developed?
Grimm: RealNetworks had developed another product called RealTimes, which is a photo sharing application that is cloud-based and the team wanted to add a feature to the product that allowed you to share photos by face. The CTO challenged our engineering team to see what they could come up with and they looked at what was available open source, off the shelf and they decided to take what was there, improve on it and built their own facial recognition feature. As part of that effort, they ended up beating the internal deadlines or milestones in the project by a matter of months and also achieved higher levels of accuracy than anyone expected. And, as a result of the incredibly diverse dataset, the team created a feature that the leadership at RealNetworks realized was not a feature in another product, it was a product and, in fact, SAFR was an extraordinarily effective and powerful technology that could be the foundation for not only a new business, but the future of this company as a result.
SIW: What would you say differentiates your facial recognition technology from others on the market?
Grimm: SAFR is the highest effective accuracy solution for live video today. We make that claim based on the combination of our high level of accuracy as evaluated by NIST testing… and our speed. What distinguishes SAFR is we are, by far, the fastest. What matters in live video is the combination of your accuracy and your speed. Because of SAFR’s speed we actually complete multiple recognitions in the time that it takes our competitors to finish one in some cases. As a result of that, we reach a higher level of accuracy, 99.9% accuracy, faster than anyone else in the world.
The second thing that sets us apart is our remarkably low level of bias. The NIST study that came out in January focused on bias, specifically the differentials in accuracy of facial recognition on individuals based on their ethnicity, skin tone or national origin. You will see that SAFR actually had the second lowest level bias algorithm in the world in that one-to-one, vendor test that NIST released.
SIW: How do you balance the privacy concerns that are so prevalent with this technology today with actionable data needed for security applications?
Grimm: Certainly, privacy concerns matter a great deal to us. Our first responsibility as a developer of this technology is to produce a truly excellent product that is, in fact, high accuracy, low bias and works in sub-optimal conditions. In addition to that, our product needs to be simple for our customers, who are system integrators, and their customers who are end-users to take into account the interests and rights of the citizens that they may be observing for very legitimate reasons. The way we do that is we ensure all the data that SAFR is collecting on individuals is encrypted in transit and at rest using bank-level encryption. We also make sure that our product includes controls so that an end-user can make sure that only the right people have access to the system to do the right functions.
SIW: How would you characterize the state of facial recognition adoption in security and how do you see that evolving?
Grimm: The customers that use SAFR are the customers that need facial recognition to work. It is critical that it identifies people on a watchlist they need to be notified of every time. Well, who are those customers? Who are the people in the world that need that level of insight and reliability? We are seeing adoption from customers in law enforcement, the gaming industry, and law enforcement in airports and transit centers around the world like train stations. Those are the customers that tend to adopt facial recognition and SAFR.
There is, of course, a flip side to SAFR’s power and any reliable facial recognition system’s power and that is it can be used to identify someone that is a VIP. If you are entering a stadium or retail store and you are a very highly valued customer, all of sudden that retail institution or stadium can provide a level of customized personal service that they wouldn’t otherwise have before.
SIW: What are some of the biggest barriers to adoption today?
Grimm: Privacy concerns do act as a drag on the adoption of facial recognition for a variety of very legitimate reasons. The greater drag on adoption on this technology is a lack of understanding of how it actually works and what it means in terms of the impact it can have on the enterprise for security, safety, and convenience. When an enterprise installs a system like SAFR, they own the data, it is their software and they control how it is being used. It’s very simple for them to train their operators on how to use it in a way that respects peoples’ rights that you’re only observing people and being notified if someone enters the space that is already on a watchlist that you already own. Facial recognition can and is being deployed in a very responsible way but unfortunately, what happens is there tends to be a level of hysteria around how the technology could possibly be abused that I think holds some people back from deploying it in a way where it could really provide a substantial improvement to the safety and security or convenience that individuals experience in the spaces we care most about.
