Technology
A decade of engineering on one hard problem.
Since 2015, Camvi has worked on a single question: how do you recognize people accurately in difficult, real-world video? Everything in the platform builds on that answer.
The engine
Built for the wild, not the lab.
Recognition in degraded conditions
Camvi's deep-learning models were engineered — and independently benchmarked — on "wild" imagery: oblique angles, partial occlusion, motion blur, and low light. That is what production cameras produce, and it is where accuracy actually matters.
Person re-identification
When a face isn't visible, appearance-based re-identification connects sightings across cameras — enabling property-wide trajectories rather than isolated detections.
Search at scale
Matching engineered for very large galleries and long video histories, with sub-second query response — so investigations feel like search, not archaeology.
A trained model suite
Detection models for weapons, objects, crowds, and incidents run alongside the identity core — and site-specific models can be trained on a customer's own footage during deployment.
Independent measurement
Benchmarked by NIST.
Camvi submits its algorithms to the U.S. National Institute of Standards and Technology's Face Recognition Vendor Test — the industry's most rigorous independent benchmark.
#1
Wild-image accuracy, NIST FRVT 1:1 verification — November 2018 report
#1
Mugshot-photo accuracy, NIST FRVT 1:1 — July 2019 report, 139 algorithms from 88 developers
NIST results are point-in-time rankings of submitted algorithms, not endorsements or certifications. Read the full benchmark history →
Privacy & Security
Powerful technology, deliberately constrained.
Recognition this capable has to be governed. Camvi's architecture makes restraint the default.
Watchlist-based by design
The system matches against lists operators explicitly enroll and govern — not open-ended identification of the general public.
Data stays with the operator
On-premises and private-cloud deployments keep footage, templates, and watchlists inside customer infrastructure, under customer retention policy.
Closed-network capable
Full functionality on air-gapped or closed networks for environments where nothing may touch the internet.
Human-in-the-loop
Camvi surfaces matches, alerts, and leads with evidence. Decisions — enrollment, response, enforcement — belong to trained people under the operator's policies.
Put the engine through its paces.
Bring difficult footage. That's the point.