Ask IREX and Video Search across Petabytes of Video

MetroCCTV, an IREX partner operating the platform in production, reports results in under a second across petabytes of video. Latency in your own deployment depends on the estate and is measured during the pilot. On a large cluster, searches are queued and executed in order, so a heavy query runs while the operator works elsewhere in the platform.

Ask IREX is shipping in beta version on selected instances, and it is included in the standard IREX price with no separate license or subscription fee. The only additional cost is a local GPU node for inference where a deployment needs one. Natural-language interpretation is best-effort: when a result looks wrong, the operator refines the wording or falls back to the on-screen event search, which stays fully available.

Yes. Facial-attribute search covers gender, age band, ethnic appearance, distinctive features such as facial hair and glasses, and an estimated spoofing probability, and appearance-based person re-identification builds a profile from clothing texture, hat and hair color, and accessories such as bags and backpacks. Where the face is not visible at all, ObjectTrack Pro calculates silhouette descriptors and matches a silhouette taken from a photo or from an existing event card against them.

Yes. Video uploaded from the field goes through the Gallery and is run through the same analytics modules that run on a live camera, so it lands in the archive as searchable events rather than as an attachment. Uploading and analyzing external video is itself a Case ID action, and access to uploaded media is granted per user group exactly as camera access is.

Only users whose role grants access to the relevant cameras, watchlists, and locations. The permission-driven interface is one of the six Ethical AI pillars: features, databases, match alerts, and search results are all scoped by role, and an agent acting for an operator inherits that operator’s permissions without any ability to escalate them.

Through the Logbook and its exports. A supervisor filters it by user, date and time, or event type, searches it by Case ID to retrieve every action taken under one case across operators and months, or uploads a photo or descriptor to find the actions that used a similar face. The results compile into a downloadable CSV report, and signed exports go daily to a secure external archive that an ethics committee, inspector general or court-appointed auditor reads without needing platform access. Actions inside the Logbook are logged too, so auditing the audit trail leaves its own entry.

It is data fusion with a defined perimeter, which is the honest version of the label. What IREX brings into one place is the video from every connected camera, the detections every analytics module produces, the watchlists an agency has lawfully enrolled, the external law-enforcement vehicle databases the platform is integrated with, and video files uploaded from the field and run through the same analytics, all searchable under one Case ID and one audit trail. What it is not is a CAD or an RMS: IREX holds no calls for service, no reports and no case records. It delivers alerts into real-time crime centers, 911 dispatch centers and secure messengers, and its alarm-monitor webhook lets any event trigger an action in a third-party system.

Only in the directions the platform documents. Outbound: low-latency alerts are delivered through real-time crime centers, 911 dispatch centers and secure messenger applications, and the generic alarm-monitor webhook pushes an event and its frame to any HTTP endpoint, so a CAD, an RMS or a notification tool can be triggered by an IREX event. Inbound: the platform integrates with external law-enforcement vehicle databases for stolen, uninsured and unregistered checks, and with external devices such as Sigur access control and Netping. Named CAD or RMS product integrations are not documented, and neither is operator single sign-on against your identity provider, so both are scoped in the design phase rather than assumed here.

Architecturally it is built for that. As a public sector data fusion platform it is multi-tenant, so several agencies share one deployment as tenants, users sit in a hierarchical catalog of groups each granted specific cameras, watchlists, locations and monitors, and every search, match and tracking action is written to an append-only log that supervisors and independent oversight bodies can read. Fusion center analytics over video is what the product does. IREX publishes no named fusion-center deployment, so judge it as regional fusion center software on the production references it does publish and on a pilot with your own archive.

No. IREX indexes what its cameras saw, under rules that keep the index narrow: only pre-registered people and vehicles are recognized, a privacy-sensitive search records a lawful basis and a Case ID before it runs, results are visible only to roles entitled to that camera, watchlist and location, and every action is logged. Criminal intelligence software in the records sense holds tips, associations, intelligence reports and case files, and IREX holds none of those. It is the video-evidence layer an investigator queries, not the system of record.