CAM-108
Leaning out over the landing rail
The block in mid-morning: the tables area on the left, the row of cells on the right and an officer standing in the foreground with his back to us. Above, on the upper landing, a person is leaning over the railing, both hands on the handrail, body out past it, looking down at the floor of the block. People are walking directly underneath. IRIS sees a body out over the rail; it does not know who it is and does not need to.
Turn on IRIS Vision
AI-generated image
What IRIS understands
What it recognises in this scene
- person over rail
- upper rail
- floor below
- officer on floor
The rule
You decide once. IRIS applies it always.
If
a person's body goes out past the handrail of the upper landing
Then
- alert the officer down in the block, with the camera and the stretch of railing
- open the live video of that stretch, so nobody goes up blind
- log the time and the stretch: the same spot coming up again and again is a problem with the spot, not with a person
The instruction that set it up«Tell me the moment somebody leans out over the upper landing rail or climbs onto it.»
What changes
What changes for whoever answers for the facility
This is not about odd behaviour or about judging anyone: it is about a height. An upper landing is a first floor, and past the handrail there is a drop of several metres onto concrete. The rule is written by the facility — which stretches, at what hours — and it is a concrete rule about a situation you can see: there is a body out over the railing. Today that gets noticed if the officer downstairs happens to look up at that moment, and a lot happens downstairs at once. With IRIS the alert reaches him with the exact stretch while it is happening. Whoever goes up, whoever talks and whoever decides what to do is still a person.
What people usually ask
Is this unusual-behaviour detection?
No, and the difference matters. IRIS does not classify behaviour, does not score anyone and does not decide what counts as normal on a block. What there is here is a concrete rule written by the facility about a concrete, visible situation: a body out past the handrail of the upper landing. Either you can see it or you cannot, and anyone watching the video can check whether the alert was right. A label like "unusual behaviour" cannot be checked, and a machine has no business judging people; this can be checked, and it judges nobody.
Does IRIS know who is at the railing?
No. It does not recognise faces, does not compare them against any database, does not store who anyone is and does not follow a particular person around the block. It sees a body somewhere the facility has said no body should be, and it raises the alert. That is why it is no use for disciplining any individual: the alert says what is happening and where, and the person who looks at the video and decides what to do is an officer.
Do you need a rule for every single thing that could happen on the block?
No, and trying would be a bad idea. A facility has three or four situations it genuinely wants to see on the cell block, and those get written one by one, in a sentence each, with their stretch and their hours. The rest is deliberately left out. A system that alerts on everything is a system nobody watches, and on a block that is worse than having nothing.
Demo · IRIS at work
You set the rule once. IRIS applies it every time.
You will see a summary of the interface, played step by step and hands-free. Each round starts with another case. The complete tool does not fit in a demo.
- The alert fires
- It lands in the tray
- You open it
- You confirm or dismiss
Next step
Who watches a whole station at rush hour?
17 cases: 11 raise an alert when something happens and 6 only measure. You see each image as it is, and then with what IRIS understands on top.
Who sees the sales that never reach the till?
10 cases: 4 raise an alert when something happens and 6 only measure. You see each image as it is, and then with what IRIS understands on top.
And the best part
A wall of monitors is impressive.Nobody can read forty tiles at once.
The more tiles there are, the less each one gets looked at. IRIS watches all four thousand cameras at once and points at a single one: the one that needs someone right now.
Tell us your problem and we will say whether IRIS understands it, or not yet.
We reply the same working day.