A Gulf industrial-zone operator
Logistics / Industrial Zones · Gulf
The shape of this problem
Too much arrives to check it all, so you sample
The problem
Operations teams needed actionable monitoring from the Al-Wakra Logistics Park video streams. The cameras were already recording, but nothing in the footage was searchable - establishing what had happened meant a person watching it. Events were therefore found after the fact if they were found at all, and the practical value of the whole camera estate was capped by however many review hours the team could spare.
What we built
We delivered a centralised computer-vision service that analysed operational footage, detected the defined events, and fed the results into review and response workflows.
What moved
| Measure | Before | After |
|---|---|---|
| Detection accuracy on the defined operational events | no prior measure | 89% under representative conditions |
| Time to locate a specific event in the footage | ~15 minutes | under 30 seconds |
| Camera hours under continuous analysis | no prior measure | 100%, previously bounded by review capacity |
Time to locate a specific event in the footage
Figures are from the practice's own delivery records for the engagement named, measured against the process that preceded it.
What it turned on
Separating recognition, OCR, localisation, and contextual enrichment made quality measurable component by component, so the weakest part could be improved without replacing the whole solution, and low-confidence detections were kept for validation rather than counted as confirmed events. Relevant to any operator holding large volumes of camera footage nobody has time to watch.
- Service line
- Inspection & Safety Vision AI
Start here
Which of the four is yours?
Tell us the documents and the monthly volume and we will send the two closest records, with the figures behind each and what it would take to repeat them on your material.
You get a reply within one working day, from the engineer who would do the work.