Platform · Facial Recognition
Identity that holds up in the wild.
Most recognition looks good on clean portraits. Camvi was benchmarked first on 'wild' imagery — bad angles, occlusion, motion, low light — the conditions real cameras actually produce.
Benchmarked, not asserted
Measured by NIST on the hardest categories.
NIST's Face Recognition Vendor Test (FRVT) is the U.S. government's own independent benchmark. In the November 2018 FRVT 1:1 verification report, Camvi ranked first for wild-image accuracy — the category closest to surveillance conditions. In the July 2019 report, Camvi ranked first in mugshot-photo accuracy among 139 algorithms submitted by 88 developers.
Rankings are point-in-time results, so we publish them with dates. What they demonstrate is durable: an engine engineered for degraded, real-world imagery rather than staged enrollment photos.
Full NIST historyHow it's used
From verification to search.
Watchlist alerting
Live matching against operator-enrolled lists — VIPs, excluded persons, known threats — with alerts routed to the right team.
Identity search
Search indexed video history by face and find every sighting — engineered for large galleries with sub-second response.
Verification
1:1 identity verification for access and enrollment workflows, on the same engine and infrastructure.
Test it on your conditions.
Bring your hardest footage — angles, lighting, crowds. That's the environment this engine was built for.