Rank One face biometric accuracy leaps in latest NIST benchmarking
Rank One Computing has reported a 30 to 40 percent leap in accuracy in the latest NIST FRVT ongoing benchmarking, achieving 99.12 percent biometric matching accuracy when the False Match Rate is set to 1 in 1,000,000.
“The accuracy of this algorithm is representative of how powerful ROC’s AI/ML face recognition has become on diverse imagery, with accuracy measured on persons from around the world,” said Scott Swann, ROC’s CEO, in a release celebrating what the company claims is unmatched biometric accuracy.
“The result of this unparalleled accuracy and efficiency is simple: there is almost no FR application that ROC’s algorithm cannot support.”
Rank One’s algorithm came in the top 20 percent of all biometrics vendors in seven of the eight benchmarks accuracy metrics and, of course, is slated to improve further. Dr. Brendan Klare, ROC’s chief scientist, indicates that “this new release is the first in a promising new line of algorithm R&D being pursued by Rank One and includes several new SDK enhancements and features.”
Another new algorithm release at the start of 2022 is poised to deliver further reductions in error rates, while a new research paper on facial recognition finds that more work across the board is needed to remove biases beyond the known demographic differentials such as for ethnicity and gender.
Rank One recently made a declaration with the Security Industry Association to campaign for a commitment to ethical facial recognition.