New algorithm pushes into biometric accuracy leaders in latest NIST FRVT

New algorithm pushes into biometric accuracy leaders in latest NIST FRVT

Deep Glint has held onto its top overall spot for biometric accuracy in the latest Face Recognition Vendor Test (FRVT) 1:1 Verification from the National Institute of Standards and Technology (NIST), but another new entry from China has leaped into the leader group.

The January 19 edition of the report shows further improvement in facial recognition accuracy by several developers.

Changes since the December 18 edition include evaluation of algorithms from four newly-submitting biometrics developers; Herta Security, Irex AI, Shenzhen University-Macau University of Science and Technology, and Vietnam Posts and Telecommunications Group. New algorithms have also been submitted by 13 developers who have previously participated in the FRVT. As a result, ten algorithms were retired from the test to limit developers to two facial recognition algorithms each.

A total of 262 algorithms from more than 150 developers were evaluated by NIST.

One of the newly-added algorithms is one from China-based Moontime Smart Technology, which placed first in both visa categories, second in both mugshot categories and the visa border category, and third and ninth in the two other border categories. The mt-003 algorithm also placed 125th in the wild category. Moontime is currently ranked 15th on the overall leaderboard.

In the December FRVT, Moontime’s top result was 12th in the ‘Border 1E-05’ category.

Sensetime took first in the two mugshot categories, VisionLabs swept the three border categories, and Beihang University-ERCACAT and Paravision had the highest accuracy against the wild and child exploitation datasets, respectively.

CyberLink notes in an announcement that its FaceMe algorithm finished sixth in the wild category, but first among vendors based outside of China, and with full global market access.

This post was updated at 8:01 am Eastern on January 29, 2021 to note CyberLink’s result in the wild category.

Related Posts

Article Topics

 |   |   |   |   |   |   |   |   |   |   | 

Comments

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