ROC lauds algorithms’ accuracy and efficiency scores in NIST facial recognition test
Rank One Computing (ROC) has achieved impressive results in the latest Face Recognition Vendor Tests (FRVT) by the National Institute of Standards and Technology (NIST), according to a company announcement.
Specifically, ROC says it achieved the best combination of accuracy and efficiency when measuring the mean FRVT rankings in all eight accuracy benchmarks: Visa MC, Visa, Mugshot, Mugshot 12+Yrs, VisaBorder, Border 10-6, Border 10-5, and Wild, respectively.
The company’s algorithm also achieved high scores in all four efficiency benchmarks: template generation speed, template size, binary size, and comparison speed.
In terms of pure accuracy metrics, the ROC SDKv2.2 algorithm ranked as the most accurate among U.S. vendors on Border crossing data (1E-5) in the one-to-one FRVT, and sixth globally.
According to a blog post on the ROC website, the latest version of the company’s biometric algorithm is 2-5x faster and 10-20x more efficient than its predecessor (v2.0), which explains its higher performance in the latest FRVT.
Additionally, v2.2 also introduced a 5x improvement in liveness fraud detection, expanded PersonID searching capabilities, and a new license plate recognition (LPR) functionality.
A notable new entry to the NIST FRVT 1:1 is from Intema-LGL Group. The Intema algorithm submitted July 15 appears among the top 6 accuracy results in the late-July update across 4 different categories. Intema acquired VisionLabs in late-2021, and the latter has not submitted an algorithm to either FRVT since then.
The Rank One results come months after two ROC executives received the SIA Industry Advocate Award for contribution to ethical development and use of biometric technologies.
More recently, the company opened an office in the Vantage Ventures accelerator offices at West Virginia University, and weeks later launched a pilot of its live facial recognition at the University campus for visitor management and threat detection.