Orbital
Satellite & survey
Wide-area imagery at metre scale. Change detection across revisits — what moved, what appeared, what’s gone since the last pass.
◉ Visual intelligence platform
Turn any visual data
into intelligence.
One pipeline takes satellite imagery, aerial and drone footage, body-cams, fixed CCTV and archived stills — and resolves every source into structured, real-time detections you can actually query.
Feeds we ingest
Satellite◇
Aerial survey◇
UAV / drone◇
Body-cam◇
Dashcam◇
Fixed CCTV◇
Thermal◇
Maritime◇
Traffic◇
PLC & SCADA◇
Industrial telemetry◇
Production line◇
The platform
One pipeline,
every altitude
The same models run from 400 km up to a camera on someone’s chest. What changes is resolution, motion and latency — not the pipeline.
Aerial
UAV & fixed-wing
Full-motion video off a moving gimbal. Tracking that survives camera motion, altitude change and handoff between airframes.
Ground
Body-cam & CCTV
Thousands of fixed and worn cameras. Detection runs at the edge, so only structured events cross the network — not raw video.
How it works
Feed in, intelligence out
Five stages between a camera and an answer. Every stage is inspectable — you can see why a detection fired, not just that it did.
-
Ingest
RTMP, RTSP, ONVIF, S3 drops and raw satellite tiles. No re-encoding upstream.
-
Normalise
Every feed lands in one frame format, colour space and clock, whatever it came from.
-
AI inference
Detection, classification, tracking and change detection — with confidence and model provenance on every result.
-
Analytics
Detections become behaviour: counts, dwell, routes, anomalies and thresholds that fire alerts.
-
Query & alert
Ask across every feed at once — live streams and the whole archive — and get told the moment it matters.
Deployment
Your premises, or ours
The same pipeline ships both ways. Where it runs is your decision, and it is reversible — not a fork in the product.
On-premise · Edge
Edge intelligence, on your site
Inference runs on hardware inside your own facility. Video never leaves the building — no footage crosses your firewall, and the system keeps working when the link to the outside world does not.
- Runs fully air-gapped
- Footage stays on your premises
- Latency measured on your LAN
- Your hardware or ours
Cloud · Managed
Managed cloud, nothing to rack
We run the pipeline and you connect feeds. Capacity scales with the number of cameras rather than with a purchase order, and model updates arrive without a maintenance window.
- No hardware to buy or maintain
- Scales with camera count
- Region of your choosing
- Continuous model updates
Most deployments end up hybrid: inference at the edge for anything latency- or privacy-sensitive, cloud for archive, search and cross-site analytics.
What changes
From visual data to decisions
Most organisations already collect far more imagery than anyone can look at — live streams, archived footage, drone sorties, satellite passes. Putting models on it turns that backlog into something that answers questions.
Nothing goes unreviewed
Every feed, sortie and satellite pass adds hours nobody will ever open. Models read all of it, continuously, and surface only what needs a human.
Cost decouples from volume
Doubling the imagery you collect normally means doubling the people who look at it. Here it means adding compute — analysis cost stops tracking data volume.
Minutes instead of days
Finding one event across an archive means reviewing by hand. Query the index instead — by object, time, place or source — and go straight to it.
Live, not retrospective
Visual data is usually consulted after the fact, to explain something. Analysis at ingest means you are told while there is still time to act on it.
Insight you never had
Counts, change over time, movement patterns and anomalies fall out of the same analysis. Decisions about resourcing and priorities stop being guesswork.
Works with what you already collect
No new sensors required. Existing streams, archives, image sets and satellite products all feed the same pipeline.
Security
Visual data is the most
sensitive material you hold
Treating it that way is a design constraint, not a compliance page.
Edge-first inference
Analysis runs next to the source. Raw imagery stays where it was captured unless you move it deliberately.
Provenance on every result
Each detection carries the model version, source camera and timestamp that produced it.
Retention you set
Frames, detections and archives expire on separate schedules, per feed.
Scoped access
Permissions attach to cameras and regions, not to a single all-or-nothing account.
◉ Get in touch
Bring us your
hardest feed.
Satellite, drone, body-cam or a decade-old CCTV rig — tell us what you already run, and we’ll come back with what your footage can actually tell you.
Contact us