PLATFORM · CAMERA MANAGEMENT
Works with the cameras you already own.
No proprietary recorder to rip out, no per-camera hardware to buy. If a camera speaks ONVIF or produces an RTSP stream on your network, Sentinel runs the same intelligence across it — face match, plate read, watchlist alerts, audit-grade evidence. From one camera to thousands, managed uniformly. Onboard the cameras you own in an afternoon.
Click to enlargeAny ONVIF camera, managed uniformly — from one to thousands.
Discover them automatically, or add one by URL.
Point the agent at the network and ONVIF discovery finds the cameras for you. Prefer to add one by hand? Paste the RTSP URL and a five-second probe checks it is real before you ever save the record.
- ONVIF WS-Discovery runs on the agent LAN segment and lists each device with maker, model, IP and supported profiles
- One-click add pre-fills the RTSP and ONVIF endpoints from a discovered device
- Manual RTSP add with optional ONVIF endpoint and credentials for PTZ
- Test-RTSP probe reports connect time, first-frame time, FPS, resolution and codec — with a friendly diagnosis when it fails
- Multi-agent aware: an agent selector decides which agent hosts each camera
Click to enlargePoint the agent at the network — ONVIF discovery finds the cameras.
From the enterprise rack to the cheap camera on the back porch.
Most platforms refuse anything but their own hardware. Sentinel meets your cameras where they are, across three honest tiers — and a Cameras Wizard so a non-technical user never types a stream path by hand.
- Tier 1 — RTSP out of the box: Dahua, Hikvision, Axis, Reolink, Amcrest and any ONVIF camera
- Tier 2 — unlockable: Tapo, Wyze, Reolink and Eufy wired models, via per-brand RTSP templates in the wizard
- Tier 3 — hostile cloud (Ring, Nest, Arlo): flagged honestly as unsupported rather than faked
- Cameras Wizard: pick a brand, follow the maker-app steps, auto-fill the URL, test the stream, see a live preview, save
- The same intelligence runs across every tier — the agent never needs to know the brand
The three tiers
What plugs in, said plainly.
RTSP out of the box
Dahua, Hikvision, Axis, Bosch, Vivotek, Uniview, Reolink, Amcrest and any ONVIF Profile S/T camera speak RTSP natively. Paste the URL — or let ONVIF discovery find it — and you are streaming in about thirty seconds.
Unlockable consumer cameras
TP-Link Tapo, Wyze, Reolink and Eufy wired models expose RTSP once a setting is toggled in the maker app. Per-brand URL templates are baked into the Cameras Wizard, so a non-technical user never types a stream path by hand.
Hostile cloud — named honestly
Ring, Nest, Arlo, Blink and HomeKit lock their video to the vendor cloud with no usable stream. We flag these up front rather than pretend support exists — so you know exactly what plugs in before you commit.
Tell Sentinel what each camera is, and where it looks.
A camera is not just a stream — it has a bearing, a field of view, a range, a purpose and a lighting condition. Capture that once and a cone of visibility appears on the map, while the intelligence layer reads every detection in the right context.
- Compass widget — drag to set the bearing; the field-of-view cone re-renders live as you edit
- Range slider with a metre readout, auto-clamped to a sensible band
- Intended use (face / plate / general / both) drives downstream filtering — plate-only cameras stay out of face-pattern math
- Detection quality auto-scored nightly from real face-match confidence; an operator override is never overwritten
- Lighting profile — indoor, outdoor-day, outdoor-night or mixed — so analytics read each camera correctly
Click to enlargeBearing, field of view, range and purpose — the cone of visibility, drawn on the map.
Inside the profile
Six fields that make every camera honest.
Bearing & field-of-view cone
Drag a compass pointer to set each camera bearing and field of view; a translucent cone of visibility re-renders live on the map as you edit, colour-coded by what the camera is for.
Effective range
A range slider with a metre readout sets how far each camera usefully sees, auto-clamped to a sane band. The cone on the map grows and shrinks with it.
Intended use
Tag a camera as face, plate, general or both. Plate-only cameras are then excluded from face-pattern analysis everywhere — the honesty contract that keeps the intelligence layer accurate.
Detection quality
A nightly job scores each camera 0 to 100 from its real face-match confidence over the trailing week. Too few samples shows calculating rather than a misleading low score, and an operator override is never overwritten.
Lighting profile
Mark a camera indoor, outdoor-day, outdoor-night or mixed so downstream analytics read its detections in the right context.
Georeferenced placement
Drop each camera on the map, drag the pin to re-georeference, and see neighbouring cameras within about a kilometre — the foundation for cross-camera tracking and coverage overlays.
Organise the network, then mark the regions that matter.
A hierarchical group tree keeps thousands of cameras navigable; zones turn a still frame into the rules an operator actually cares about — crowds, loitering, lines crossed, areas that should stay empty.
- Hierarchical groups — for example a city, then a district, then a square — with drag-to-group and a group filter
- Groups drive RBAC scoping: an operator restricted to a group sees only its cameras
- Zones drawn as polygons over a still frame: crowd-counting, loitering, line-crossing, restricted-area
- Per-zone thresholds — occupancy warning and critical levels, loitering dwell time, line-crossing direction
- Zone events arrive in the feed tagged with the zone name
Click to enlargeA city, a district, a square — thousands of cameras stay navigable.
How it works
Connect, profile, zone, go.
Connect
Point the on-site agent at your network. ONVIF auto-discovery lists every camera it finds with maker, model and IP; or paste an RTSP URL and run a five-second test probe that reports connect time, FPS, resolution and codec before anything is saved.
Profile
Set the Camera Profile — bearing, field of view, range, intended use and lighting — and watch the cone of visibility appear on the map. Tune per-camera detection toggles and confidence thresholds against a sample frame.
Zone
Organise cameras into a hierarchical group tree, then draw zones over a still frame — crowd-counting, loitering, line-crossing or restricted-area — each with its own thresholds. Zone events arrive tagged with the zone name.
Go
The camera goes live with its AI overlay, snapshots, hover-preview and PTZ where supported. Its DPIA, privacy notice and retention window travel with it — accountable from the first frame.
Under the hood
Specifications
| Camera connection | Any ONVIF Profile S/T camera, or any source that exposes an RTSP URL on the LAN |
| Discovery | Agent-side ONVIF WS-Discovery; discovered devices list maker, model, IP, port and supported profiles |
| Manual add | Paste RTSP URL + optional ONVIF endpoint and credentials for PTZ; assign group, location and map pin |
| Test-RTSP probe | Five-second probe returns connect time, first-frame time, FPS, resolution, codec and audio-track presence before the camera is persisted |
| Consumer-camera tiers | Tier 1 RTSP out-of-box · Tier 2 unlockable via per-brand templates · Tier 3 hostile cloud, flagged honestly |
| Camera groups | Hierarchical parent-to-child tree; drag cameras between groups; group filter and RBAC scoping |
| Zones | Polygon regions of interest — crowd-counting, loitering, line-crossing, restricted-area — with per-zone thresholds |
| Camera Profile | bearing (0-360) · field of view (1-360) · range (metres) · intended use (face / plate / general / both) · lighting · detection quality (0-100) |
| FOV cone overlay | Translucent sector drawn from bearing / FOV / range, colour-coded by intended use, live-updating on the map |
| Multi-agent | Each camera is bound to one agent; an agent selector decides which agent hosts each discovered camera |
| Scale | List view handles 5,000+ cameras with search, group filter, sort and pagination — never an unbounded render |
| Per-camera governance | DPIA record, auto-generated localised privacy notice, and per-camera / per-zone retention windows |
Specifications describe shipped platform capabilities; we'll confirm the configuration that fits your deployment during your demo.
The bigger picture
Cameras are the front door to the whole platform.
Live Operations
Once cameras are on, they render in the Command Center map and the multi-camera Live View grid, each with its live AI overlay.
Learn moreThe On-Site Agent
A single binary at the site does the discovery, the RTSP-to-HLS streaming, the on-device AI, and the PTZ control — all locally, on CPU.
Learn moreAI Vision Pipeline
Each camera optionally runs face, plate, person, vehicle and behaviour detection through the agent or the worker pool.
Learn moreBring your own cameras. See them live in an afternoon.
Request demo access and we'll send you a private, pre-loaded environment. Discover a camera over ONVIF, paste an RTSP URL and watch the test probe pass, set a Camera Profile and see the cone of visibility appear on the map — on real cameras, not a slide.