Newer algorithms join incumbent facial recognition developers among accuracy leaders
A series of new facial recognition algorithms submitted to the U.S. National Institute of Standards and Technology for its Face Recognition Vendor Test 1:N Identification appear on the leaderboard as of the July 5 update.
NIST found algorithms from Megvii, Paravision, Kakao, Armatura, Verigram, Maxvision and Omnigarde, each submitted within the last three months, placed them within the top 22 vendors in overall biometric accuracy.
Paravision scored the second-most accurate result in the Visa-Kiosk category, along with top 5 finishes in the categories for mugshots with both databases of 12 million and 1.6 million images, matching webcam probe images to mugshots, the Visa-Border category and the one for mugshots after 12-plus years. Paravision also ranked in the top 10 for border images compared after 10 years or more.
“The recognition of Paravision face recognition as the top performer in the U.S., UK, and EU in both NIST FRVT 1:1 and 1:N and across so many different benchmarks underscores the wide-ranging usefulness of our Gen 6 technology,” comments Paravision CTO Charlie Rice. “We look forward to bringing this NIST-benchmarked software to our partners worldwide, enabling them to deploy advanced and trusted solutions.”
The company notes a steep drop-off after the top 5 performing vendors in the Visa-Border test, and that NIST uses this category as its primary sorting metric for the accuracy leaderboard.
Paravision’s Gen 6 facial recognition software is scheduled for delivery to partners this summer.
Omnigarde also touts its position amongst U.S. developers in an announcement, as it works towards a leading position as a core biometric technology provider.
More than 350 facial recognition algorithms were tested for NIST’s July 5 FRVT report.