ROC is claiming the top spot overall among all American biometrics developers in NIST’s facial recognition evaluations. A new algorithm from the company delivered accuracy, efficiency and speed results among the most accurate ever, according to a company announcement.
The latest 1:N Identification report of the Face Recognition Vendor Test (FRTE) was published on August 27, and the 1:1 Verification evaluation was updated on August 13.
In Investigation Search mode, matching photos from immigration lanes and registered traveller kiosks with a database of 1.6 million faces, ROC scored a Rank-1 miss rate of just below 0.001 percent. NEC and Paravision followed close behind, as the best of the rest from the West.
ROC also topped Western facial recognition vendors by false non-match rate (FNMR) in the Visa-Border and Border-Border categories of the 1:1 Verification evaluation, just ahead of Innovatrics and Paravision.
In the assessments of search speed for 12 million images and accuracy with 90-degree profile images in the 1:N Identification test, ROC came second among Western-based providers, just behind NEC.
“What sets this algorithm apart is its consistency,” says ROC Co-founder and Chief Scientist Dr. Brendan Klare. “It performs at the highest level across verification, identification, and investigative search. This consistency is what makes it among the world’s best, and why our partners trust ROC for their missions.”
Director of Machine Learning Keyur Patel attributes the algorithms’ performance to the company’s unique development process.
“ROC algorithms are built on a foundation of decades of facial recognition expertise, continually advanced with the latest in AI research and optimization,” he says. “That’s how we deliver solutions that perform with high accuracy across every scenario, from one-to-one verification to large-scale investigative searches.”
Article Topics
accuracy | algorithms | biometric testing | biometrics | Face Recognition Technology Evaluation (FRTE) | facial recognition | NIST | ROC