Technology · NIST Benchmarks
Measured by the toughest referee in the industry.
The U.S. National Institute of Standards and Technology runs the Face Recognition Vendor Test — an independent, ongoing benchmark of algorithms from developers worldwide. Camvi's results below are stated exactly as achieved: with dates and categories.
NOVEMBER 2018
#1 — Wild-image accuracy
In the NIST FRVT 1:1 verification report of November 2018, Camvi's algorithm ranked first for accuracy on the WILD image dataset — photojournalism-style images with difficult angles, occlusion, and lighting. Among the categories NIST tests, wild imagery is the closest to real surveillance conditions.
JULY 2019
#1 — Mugshot accuracy
In the NIST FRVT 1:1 verification report released July 2019, Camvi ranked first in mugshot-photo accuracy among 139 algorithms submitted by 88 developers worldwide.
ONGOING
Continuous participation
Camvi continues to develop against the FRVT methodology. NIST publishes current results for all participants at pages.nist.gov/frvt — we encourage buyers to read the source reports directly.
How to read NIST results
FRVT rankings are point-in-time results for specific submitted algorithms in specific categories — not certifications, endorsements, or permanent standings. That is why every Camvi claim above carries a date and a category. What the results demonstrate durably is engineering priority: Camvi has repeatedly performed at the top of the categories closest to hard, real-world conditions.
Accuracy claims should survive scrutiny.
Ask us anything about methodology — or better, test the engine on your own footage.