Scrubbing footage is the real cost
A pallet is short. A customer says they were here at four. A forklift clipped a rack and nobody reported it. The footage exists, and that is the problem: somebody now spends an afternoon dragging a slider across eight cameras, guessing at a time window, watching an empty yard at double speed. That cost is invisible because it falls on whoever is free, and it is why most recorded footage is never reviewed at all.
How it works
- Frames are pulled from the recorder and analysed. Rather than storing a second copy of your video, the system reads frames from your existing NVR, works out what is in them, and keeps the result.
- Results are written as events. Each event carries a time, a camera and labels for what was found. A row like "person, loading dock, 02:14" is something you can search. A video file is not.
- A short clip is captured around each event. Roughly ten seconds before and ten seconds after, so you see the approach and the walk-away rather than one still. The lead-in is usually the useful part, because it shows where somebody came from.
- You search by label and time. Filter the timeline to a camera, a period, or a label, and read what happened as a list. The labels can be taught to match your site, so you search for the words your team actually uses.
- Budgets keep analysis predictable. Per-camera and hourly limits control how much is processed across a lot of cameras.
Your footage stays on your recorder. The analysis runs on our own server rather than a third-party cloud. AI analytics sets out the whole platform.
What you see and get
A chronological list of events per camera and across the site, filterable by label and time, with a playable clip attached to each one. In practice a question that used to be an afternoon becomes a filter, a glance at the list, and a twenty-second clip. The same timeline is where after-hours alerts land, where line and zone crossings are recorded, and where stock movement checks look for events to cross-check.
Where it fits best
Anywhere a dispute, a shortage or a damaged item has to be reconciled with a time: warehouses and despatch, retail back-of-house, workshops, car yards, and sites where contractors come and go. Our own warehouse is the clearest example, with the timeline built from seven cameras on a Dahua recorder and used by us every day.
What it does not do
- It is not a full index of every frame. The timeline holds the events that analysis produced, so quiet stretches and anything missed between analysed frames will not appear as rows.
- Clips are short by design. For a long sequence you still go to the recorder, and the timeline tells you exactly where to look.
- It does not identify people. Events describe what was in the frame, not who.
- It does not decide what matters. A row saying goods were moved is not a finding of wrongdoing.
- It does not extend how long your footage is kept. Retention is still set by your recorder and its storage.