- Manufacturing
- Updated
Computer Vision for PPE Compliance: 100K+ Safety Issues Detected
The manufacturer struggled to enforce safety norms through manual PPE checks that were inconsistent and could not scale to 24/7 operations. We built a computer-vision platform that detects masks, gloves, helmets and jackets and records violations around the clock. It detected more than 100,000 safety issues, improved safety compliance 53% and cut safety-monitoring cost 30%.
Book a 30-Minute Working Session
The situation
Manual PPE checks were inconsistent and could not scale.
The manufacturer faced major issues adhering to safety norms through manual detection of safety equipment, and wanted to automate the entire process. Manual PPE checks were inconsistent and could not scale to 24/7 industrial operations. Once manual inspection was confirmed not to be accurate enough, the need was clear: a real-time, automated system that could take precautions against injuries and accidents before they happened.
What we built
Computer vision checking PPE on site cameras, around the clock.
We implemented a PPE-detection platform using AI and computer vision. Site camera feeds run into a computer-vision model around the clock, and an event-triggering system detects masks, gloves, helmets and jackets, records violations and fires real-time violation alerts, so the violation database provides inputs for injury and accident prevention. The solution combines deep learning, machine learning, computer vision and IoT on the Azure data stack, managed on Kubernetes, with image-processing algorithms automating the entire PPE-tracking process. A visualization layer feeds centralized safety dashboards on web, mobile and tablet, and the platform is designed to deploy from one factory to many.
How it works
The architecture, stage by stage.
Capture
- Site cameras, 24/7 feeds
Detect
- Computer-vision model for PPE detection (masks, gloves, helmets, jackets), using deep learning and image-processing algorithms
Alert & act
- Event-triggered real-time violation alerts and violation tracking
Dashboards
- Centralized safety-analytics dashboards on web, mobile and tablet
What changed
The outcomes, as the client measures them.
- The platform detected more than 100,000 safety issues.
- Safety compliance improved 53 percent.
- 60 percent of the time spent on manual checks was saved.
- Safety-monitoring cost fell 30 percent.
- Detection and violation alerts run in real time, 24/7.
- Safety monitoring moved from a manual, after-the-fact task to real-time, automated detection.
By the numbers.
100K+
Safety issues detected
53%
Improvement in safety compliance
60%
Less time spent on manual checks
30%
Lower safety-monitoring cost
Facing the same problem?
Start with a working session.
Bring your version of it to a 30-minute working session. We will walk through how this was built and what it would take on your data.