FB pixel

Rank One face biometrics above 90 percent accuracy with masks with multiple cameras in DHS Rally

Algorithm with higher efficiency than competitors scores high marks for performance

airport facial recognition masks

Rank One Computing (ROC) has emerged from the 2020 DHS Biometric Technology Rally among the top performers in face biometric accuracy, according to a company announcement that says it is the only vendor to supply an algorithm to achieve accuracy with masked faces of above 90 percent in combination with two different image acquisition systems.

The Rally, hosted by DHS’ Science and Technology Directorate (S&T) and carried out at the agency’s Maryland Test Facility (MdTF), was expanded from the previous year to include assessment of biometric algorithms with masked faces, and also assessed accuracy for unmasked faces. Its completion was delayed until early 2021 due to the pandemic.

ROC was assessed under the alias “Owens” for unmasked faces and “Pond” for masked faces, and achieved the second-highest overall accuracy for masked faces, at 91.9 percent. On unmasked faces, ROC scored 99 percent accuracy with three of the six acquisition systems, and 94 percent with four out of six, and was one of only two to achieve that level of accuracy, according to a blog post by ROC Founder, President and Chief Scientist Dr. Brendan Klare.

“In addition to the accuracies achieved by the ROC algorithm, it continues to be a stalwart of efficiency in the NIST FRVT benchmarks,” writes Klare in the post. “The combination of top-tier accuracy and hardware efficiency both put Rank One in a category of its own and save ROC’s partners and customers substantial amounts of money in hardware costs.”

The company also compares the efficiency of its algorithm to leading algorithms from competitors in NIST FRVT results, finding its algorithm between three and seven times more efficient in terms of hardware resource usage. ROC algorithms have been found by NIST to generate templates in under 200 milliseconds, and generate templates of only 261 bytes, with an algorithm footprint of 83 megabytes for one-to-one matching and 135 megabytes for 1:N.

In addition to 10 facial recognition algorithms, three iris biometric matchers were tested, all of which delivered lower accuracy than ROC.

Rank One appointed former Idemia NSS head Scott Swan as its new CEO, with Klare moving over into his new positions, at the beginning of the year.

Related Posts

Article Topics

 |   |   |   |   |   |   |   |   |   | 

Latest Biometrics News


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Most Read This Week

Featured Company

Biometrics Research

Biometrics White Papers

Biometrics Events

Explaining Biometrics