In a recent LinkedIn post, Yoti CEO Robin Tombs announced significant advancements in Yoti’s Facial Age Estimation (FAE) technology, with enhanced accuracy across various age groups.
The latest FAE model, launched in July 2024, shows a 7.8 percent accuracy improvement for 18-year-olds, reducing the Mean Absolute Error (MAE) from 1.35 in the March 2023 model to 1.25.
The technology also displays a 6.5 percent accuracy increase for 15-17 year-olds, with the MAE dropping from 1.09 to 1.02. For 10-12 year-olds, accuracy improved by 4.1 percent, with the MAE decreasing from 1.16 to 1.11. Overall, the weighted average MAE for individuals aged six through 70 has fallen by 11.3 percent, from 2.88 to 2.56.
Despite recent criticism from the CEO of Ofcom – the UK’s telecom regulator – regarding the effectiveness of FAE for children, Tombs notes that Yoti’s data shows the technology has maintained lower MAEs for younger age groups compared to 18-year-olds.
The company’s FAE has been independently tested and certified by the Age Check Certification Scheme (ACCS) since November 2020, confirming that at least 98.89 percent of 18-year-olds are accurately estimated to be under 25 years of age.
Tombs also points out that the new July 2024 model will seek independent certification from ACCS and submission to the National Institute of Standards and Technology (NIST).
Yoti continues to invest in its FAE technology to ensure it serves as an effective age assurance method for various age checks required by businesses and regulators. The technology is crucial for accurately verifying ages for purposes such as under/over 18, under/over 13, and other age-restricted product discounts or regulatory requirements.
Earlier this summer, Yoti’s “Digital ID Connect” apps were accredited under the UK’s Proof of Age Standards Scheme (PASS), endorsed by the Home Office, CTSI, and NPCC. That accreditation ensures that the Post Office EasyID, Lloyds Bank Smart ID, and Yoti ID apps meet rigorous standards, offering businesses confidence in their legitimacy.
–
(Originally published on FindBiometrics)
Follow Us