FRVT-leading facial recognition algorithm powers Paravision’s expanded product suite
The facial recognition algorithm that placed Paravision atop the leaderboard in National Institute of Standards and Technology testing is powering the new and expanded fifth generation of the company’s biometric software suite, the company has revealed.
The new generation product suite gives Paravision’s partners the broadest-yet range of capabilities and deployment options, with an algorithm topping not only other U.S. providers in the NIST Face Recognition Vendor Test (FRVT) 1:N test, but also leading developers from around the world, Paravision CPO Joey Pritikin tells Biometric Update in an interview, citing the January 24 update.
Paravision’s new algorithm cut error rates by 70 percent, according to the NIST test, and the company achieved the lowest error rates observed in multiple categories, along with top-10 results in two others.
The inclusion of this NIST-tested algorithm in the new version improves the product suite’s accuracy substantially, therefore. Pritikin notes that many companies submit algorithms not yet available in their products, or multiple algorithms, which can make it challenging for customers to know what degree of biometric accuracy they are receiving. The new version also improves deployment flexibility and speed, according to the announcement.
Paravision President and COO Benji Hutchinson emphasizes in the same interview the importance of NIST’s independent testing, and that Paravision is the only U.S.-based company to consistently achieve accuracy results among the world’s leaders, including those with the backing of the Chinese government.
The presence of a smaller company (Paravision is still around 50 people, Pritikin says) among the multinational giants and state-backed leaders in NIST FRVT testing is a testament to the team’s engineering talent. Hutchinson suggests that it also reflects a trend of smaller, more nimble companies introducing new technologies to innovate face biometrics.
Face biometrics’ place in the new normal
The fifth-generation of the company’s facial recognition product suite features new mobile SDKs including iOS support, and updates to the Edge AI SDK for Ambarella video processors.
The expanded deployment capabilities are a reflection of the spread of facial recognition into different markets and use cases, Hutchinson explains. Paravision’s customer base has grown from around 10 customers to several dozen over the past two years.
“And the majority of our customers are not in the United States, a lot of them are global,” he reveals. “They other thing is some of the big winners inside of our customer roster are from that IDV, eKYC world, and they are the leaders as far as generating our revenue and where we see growth.”
This is despite an impression within Paravision that banks are lagging behind in the adoption of biometrics for remote transactions, something Hutchinson says the company sees changing over the next year or two.
As use cases for the company’s biometric technologies expand, it has become increasingly necessary to support both iOS and Android, in addition to the various other platforms Paravision’s partners and customers use.
Android is used by customers in time and attendance systems or mobile enrollment systems for remote enrollment to civil ID systems, Pritikin explains. With the company’s increasing customer base in identity verification and KYC applications, serving the other dominant mobile platform is a natural extension of Paravision’s capabilities.
Building blocks for different architectures
“We actually serve a lot of different verticals,” Pritikin says. “This product is designed to serve a wide variety of applications and verticals with a core set of technologies that are broadly deployable.”
He mentions government services and air travel, access control, identity verification and digital KYC processes.
“At the end of the day they’re all looking for the same face recognition pipeline. It just has to fit their processor and OS and workflow and all of that, and that’s what you’re seeing from us.”
This is what motivates the modular architecture, allowing for deployment of capabilities at the edge, in the cloud, or on mobile devices.
The Ambarella optimization brings a new version of Paravision’s liveness detection to edge devices, providing biometric presentation attack detection against sophisticated attack methods like high definition videos and 3D masks.
“We’ve become a unique partner in the biometrics and identity industry because we combine world-class accuracy developed in North America, deployed in a way that’s partner-first,” Pritikin says. “Those three characteristics make up a pretty unique player in the market, and a partner of choice.”
Pritikin summarizes that it amounts to providing “an even broader set of capabilities to a broader set of use cases,” with the most accurate algorithm on the market, built in the USA.
Paravision’s Gen 5 facial recognition product suite is now available to partners worldwide.
Ambarella | biometric liveness detection | biometrics | biometrics at the edge | face biometrics | Face Recognition Vendor Test (FRVT) | facial verification | identity verification | KYC | mobile biometrics | NIST | Paravision | SDK