FB pixel

Latest NIST age estimation test shows gains by Cognitec, Idemia, Innovatrics, Yoti

‘Further evidence that estimation has a role to play alongside verification
Latest NIST age estimation test shows gains by Cognitec, Idemia, Innovatrics, Yoti
 

The National Institute of Standards and Technology (NIST) has posted fresh test results on its benchmark for facial age estimation technologies. Added in the new report are algorithms from Cognitec, Idemia, Innovatrics and Yoti

Other biometrics providers submitting new algorithms to the NIST Face Analysis Technology Evaluation (FATE): Age Estimation and Verification this year include Dermalog and Daon.

Yoti scored the lowest Mean Absolute Error (MAE) for ages 13 to 16, while Idemia’s algorithm had the lowest False Positive Rate (FPR) in the Challenge 25 task’s “application” and “border” categories. Cognitec landed in the top 5 in the Challenge 25 “application,” Age 13-16 MAE and Ages 8-12 FPR categories. Innovatrics sits 6th and 4th in the two Challenge 25 categories, and 7th in FPR for Ages 17-22.

An announcement from Dresden, Germany-based Cognitec says that in the latest evaluation, “averaging the rankings across two genders and all six regions, Cognitec achieves the best rank overall.”

“Ordering the algorithms by the average of the Mean Absolute Error (MAE) rates over all genders and regions also puts Cognitec’s algorithm in the top position.”

Yoti, meanwhile, claims success in the Child Online Safety rankings, where its algorithm occupies the top spot for MAE at 2.0182. Writing on LinkedIn, Yoti CEO Robin Tombs notes the significance of their score for 13-16 years olds: “NIST publish our 13-16 MAE as 2.06 years, significantly ahead of most other vendors. The 10th best vendor has an MAE of 3.86 years.”

In a comment quoting Tombs’ post, the Age Verification Providers Association (AVPA) acknowledges the value of the NIST rankings, but also their limits. “Given the test data is not the same quality as is used in real-world implementations, as Robin Tombs points out below, the absolute values for margins of error are somewhat misleading. Typically, in practice results are even more accurate.” 

“This is further evidence that estimation has a role to play alongside verification.”

Other familiar names hovering near the top of the NIST charts, which were last updated in February 2025, include Ondato, Incode, Nominder and ROC

Related Posts

Article Topics

 |   |   |   |   |   |   |   |   | 

Latest Biometrics News

 

Face biometrics use cases outnumbered only by important considerations

With face biometrics now used regularly in many different sectors and areas of life, stakeholders are asking questions about a…

 

Biometric Update Podcast explores identification at scale using browser fingerprinting

“Browser fingerprinting is this idea that modern browsers are so complex.” So says Valentin Vasilyev, Chief Technology Officer of Fingerprint,…

 

Passkeys now pervasive but passwords persist in enterprise authentication

Passkeys are here; now about those passwords. Specifically, passkeys are now prevalent in the enterprise, the FIDO Alliance says, with…

 

Pornhub returns to UK, but only for iOS users who verify age with Apple

In the UK, “wanker” is not typically a term of endearment. However, the case may be different for Pornhub, which…

 

Europol operated ‘shadow’ IT systems without data safeguards: Report

Europol has operated secret data analysis platforms containing large amounts of personal information, such as identity documents, without the security…

 

EU pushes AI Act deadlines for high-risk systems, including biometrics

The EU has reached a provisional agreement on changes to the AI Act that postpone rules on high-risk AI systems,…

Comments

Leave a Reply

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

Biometric Market Analysis and Buyer's Guides

Most Viewed This Week

Featured Company

Biometrics Insight, Opinion

Digital ID In-Depth

Biometrics White Papers

Biometrics Events