A wall of tiles is not monitoring
Put sixteen cameras on a screen and attention lasts a few minutes. Movement in a corner tile at quarter size, behind racking, is easy to miss. Most people watch one or two cameras properly and treat the rest as wallpaper.
Live detection does not replace the person at the screen. It marks up what the screen is already showing, so the person is reading labels instead of scanning pixels.
How it works
- It runs on the local restream. The dashboard already pulls each camera through a restream on your network rather than opening a direct connection to the recorder for every viewer. Detection reads that same restream, so three people watching one camera does not mean three extra streams off the NVR.
- Frames are analysed as they play. An analysed frame comes back with the things found in it and where they sit in the picture, and the dashboard draws a box and a label over the video.
- It labels what is there. People, vehicles and equipment are the common ones. The system can also be taught the labels that matter on your site, so they read in your language rather than a generic category.
- Budgets apply here too. Per-camera and hourly analysis budgets cap how much processing a live session can consume, so leaving a view open on a busy camera cannot quietly run up cost.
Footage stays on your recorder. Analysis runs on our own server, not a third-party cloud. AI analytics covers the platform as a whole, and how it works follows a frame from camera to dashboard.
What "while you watch" means in practice
This is live assistance, not a background service. Detection happens while a live view is open, and stops when the tab is closed. Nothing is being watched on your behalf when nobody is logged in.
That is the honest split. Live detection is for the moments somebody is looking: a delivery arriving, a shift changeover, a dock running behind. For the hours nobody is at a screen, after-hours alerts do the watching and the event timeline keeps the record.
What you see and get
The normal live grid or single-camera view, with boxes and labels drawn over it as frames come back. Useful when somebody has been called to a camera and needs to find the person or vehicle in a busy frame quickly.
Where it fits best
Front counters and gatehouses, loading docks during a shift, workshop floors, car yards, and any site where a manager checks cameras from a desk. See warehouses and logistics.
What it does not do
- It does not watch when you are not watching. Close the tab and live detection stops.
- It does not recognise faces or identify individuals, and it does not read number plates.
- It does not know intent. A person outlined on screen is a person, not a thief. The label is a fact about the picture, and the judgement stays with your staff.
- It can miss or mislabel things. Poor light, rain on the dome, heavy glare or a partly hidden body all reduce what a frame will give up.