Yoti trumpets NIST age estimation results and testing plans

A new facial age estimation algorithm submitted by Yoti to the U.S. National Institute of Standards and Technology has placed first and second for accuracy in a couple of key categories.

NIST updated its Face Analysis Technology Evaluation (FATE) Age Estimation & Verification on October 7.

Yoti scored the lowest weighted average mean absolute error (WAMAE), at 1.88, and true positive rate (TPR), at 0.57, for children 13 to 16 years old with the algorithm it submitted on September 24. The same algorithm also placed third-best in false positive rate (FPR) for 8 to 12 year-olds at 0.083, and for FPR in the Challenge 25 scenario with the better quality “Application” data set (0.072±0.018), and fourth with the medium-quality “Border” dataset (0.190±0.029).

The company was among the original crop of age estimation providers to have their technology assessed by NIST earlier this year, and also scored well in child online safety in that test.

NIST has tested 13 algorithms from 11 vendors, and Yoti Co-founder and CEO Robin Tombs notes eight other companies that have developed face biometric age estimation in a LinkedIn post. He also shared internal test data.

Tombs revealed in the post that the UK Age Check Certification Scheme (ACCS) is also now testing the company’s biometrics-based age estimation with data from 18-year-olds captured on mobile phones. Tombs says Yoti will also request a test for children aged 6 to 17 as soon as ACCS has consented test data.

The company also delved into the relationship between true and false positive rate and true and false negative rate in a recent blog post on the nuance invited by the question: “How accurate is it?”

Facial age estimation the choice of roughly 2 in 3

Yoti Chief Policy and Regulatory Officer Julie Dawson discussed policy trends, how different forms of age assurance differ from each other and why the data involved in facial age estimation does not qualify as biometric data under GDPR with DIACC President Joni Brennan in a new episode of the Trust Talks & Digital Dives podcast.

Dawson emphasizes how transparency, such as that provided by Yoti’s self-testing and NIST’s evaluations, contributes to trust in the technology.

When people are given a choice between several age assurance options, 65 percent or more tend to choose facial age estimation, Dawson says.

Article Topics

accuracy  |  age estimation  |  age verification  |  biometric testing  |  Face Analysis Technology Evaluation (FATE)  |  face biometrics  |  NIST  |  Yoti