The problem
A national construction security group monitors thousands of cameras across active sites. Streaming everything to the cloud for analysis was too slow and too expensive, and operators could not watch every feed.
What we built
Perception models optimised to run on the cameras and edge hardware already on site. Intruders and safety violations are detected in real time where the footage is captured, and only what matters is surfaced to operators.
Outcome
- 4,000+ cameras running edge AI live across the group's sites
- Intruders and safety violations detected in real time — risk reduction and cost efficiency
- Detection at the edge, with cloud used for coordination rather than inference
Client details available under NDA. Talk to us about vision AI on cameras, robots, drones, and wearables.