Computer Vision
Smart PPE Detector
A real-time personal protective equipment detection system deployed on edge hardware across industrial construction sites.
- Industry
- Construction and industrial
- Status
- Live in production
- Outcome
- 97% detection accuracy in live production
Overview
An edge-deployed vision system that monitors protective equipment compliance on active industrial sites without sending video to the cloud.
The challenge
Safety monitoring depended on periodic manual observation, so lapses were recorded after the fact rather than while work was in progress.
Architecture
- YOLOv8 detection models tuned on site-specific conditions
- OpenCV pipeline handling camera streams and frame pre-processing
- NVIDIA Jetson edge devices performing inference on site
- Event reporting to a central dashboard with retained detections only
Capabilities delivered
- Real-time PPE detection on live camera feeds
- On-device inference for constrained connectivity
- Site and camera level compliance reporting
- Alerting into existing safety workflows
Outcome
97% detection accuracy in live production.
More work