Rank One and Panasonic upgrade biometric facial recognition offerings for high accuracy

Rank One Computing has launched a new facial recognition algorithm as part of an update to its biometric ROC SDK, which the company says reduces both error rates and template generation time by half.

While the biometric algorithm included in the ROC SDK version 1.22 significantly improves speed and accuracy, Rank One points out that the version included in its version 1.20 has placed well in NIST FRVT rankings for accuracy and efficiency.

The FR Fast algorithm, which is designed for mobile and embedded applications, has also been improved with a 10 percent reduction in template generation time and reduced error rates of between 15 and 40 percent.

The v1.22 release also includes Rank One’s new periocular algorithm for identifying people wearing masks, and the company’s tattoo recognition algorithm. Several other API enhancements have been added as well, including a new database infrastructure for media management.

”ROC SDK version 1.22 represents the culmination of more than a year’s research. For the first time, we’ve been able to deliver not only significant accuracy improvements but also significant speed improvements in our face recognition pipeline – a noteworthy achievement as we already had one of the fastest face recognition algorithms on the market,” comments Rank One Chief Engineer Scott Klum. “With a new version of ROC FR Fast and the introduction of a periocular face recognition algorithm, v1.22 is an essential upgrade for existing users of the ROC SDK and a compelling alternative for users of competing products looking to improve the accuracy and efficiency of their face recognition offerings.”

Panasonic discusses FacePRO upgrades

The FacePRO biometric facial recognition system from Panasonic now provides high accuracy even with faces occluded by sunglasses or masks, and the company says it is “now on the offensive” in the search for market share.

In an internal interview with company facial recognition developers Hiromichi Sotodate, Masashige Tsuneno, and Yuiko Takase, the evolution and capabilities of FacePro are discussed. The system can perform real-time recognition on up to 2,000 cameras and 100 servers, and system engineers working at client sites helped Panasonic add functions and make the system easier to use.

FacePRO is a deep learning server software for integration with intelligent surveillance cameras, also provided by Panasonic.

Tsuneno notes the importance of minimizing the burden on the back-end server while enhancing facial recognition accuracy.

“Subjects are continually moving in surveillance camera footage. Furthermore, the cameras are installed in places that are difficult to spot, such as ceilings,” he observes. “We performed repeated error identification and problem setting, and investigated the level of robust control that could be achieved. This means control that maintains stability even if the dynamic characteristics of the control target change slightly.”

In a trial at Haneda Airport, the technology was used to check more than 60,000 people a day for two weeks, running 24 hours a day, and proved highly accurate, producing almost no false matches, Takase says.

The post also includes interviews with company system engineers, who discuss how FacePRO works in challenging lighting conditions, and the challenges of implementing facial recognition systems in different circumstances to meet high customer expectations.

The biometric engine was jointly developed by Panasonic’s Innovation Center and the National University of Singapore, and impressed in a recent demonstration to potential U.S. customers, according to the post.

Panasonic showed off FacePRO technology at last year’s ISC East.

Article Topics

accuracy  |  biometrics  |  biometrics research  |  facial recognition  |  Panasonic  |  periocular biometrics  |  ROC  |  video surveillance