Video surveillance and AI vision
Rutba Vision, for camera networks that outgrew one recorder
Each site runs on its own nodes and keeps its own footage, and a hub at head office brings the sites into one picture. Every camera is pulled once and recorded without being decoded. Detection and AI are switched on camera by camera, so you pay for inference only where it earns its keep.
In one line each
- The unit
- A node, taking the core, edge and analysis roles
- Sites
- Stand alone, or uplink to a hub
- Footage
- Stays where it was recorded
- Recording
- Never decoded, every segment hashed
One machine can take all three roles. A large site spreads them over several.
Watch
Rutba Vision on film
A short film of how a site records, what the AI layer adds on the cameras that need it, and how head office searches across sites.
The problem
Camera systems rarely fail at ten cameras. They fail at a hundred and fifty.
A hundred and fifty cameras write terabytes a day, and the failures are the same everywhere. Rutba Vision is built around them.
The disks fill and recording quietly stops
Each edge node runs a storage guard that enforces retention and free space itself, and forecasts how many days the free space buys from what each camera actually writes.
The event database grows until searches time out
Metadata is kept one file per day. Expiring a day is deleting its file, so nothing grows without end.
A clock drifts and the footage stops counting
Video is stamped in UTC as it arrives, whatever the camera thinks the time is. Camera clocks are still checked four times a day and put right.
One box is the whole system
Cameras are assigned to nodes, not tied to one. A node that goes quiet for 15 seconds loses its cameras to the others, and recording carries on there.
Every change is forty clicks nobody can review
Cameras are declared in one file. It is validated and diffed against what is running before it is applied, with no restarts.
The analytics fall over and take recording with them
Nothing above the recorder can stop the recorder. With detection, AI and search all down, footage is still written, catalogued and kept.
What is in it
Where to read next
How it works
Sites, nodes and the hub, and why each camera is pulled only once.
Viewing
Live views built on the wall itself, and a timeline that shows where things happened.
Analytics per camera
Zones, masks, object filters and rules set for each camera on its own.
Search
Ask for “red car at the north gate after 10pm” and get the footage.
Cameras and setup
One file for every camera, ONVIF discovery and clocks kept in line.
Running unattended
Retention, repair and upkeep on day two hundred, with nobody watching.
Security and evidence
Hashed segments, export manifests, holds and an audit log.
Ways to deploy
On your site, managed by us, or many sites under one hub.
How many cameras are you running?
Tell us the sites, the camera counts and what you need to detect. We will come back with a sized set of nodes, a retention figure and a straight answer on which analytics are worth switching on.