PPE Compliance Detection: AI-driven real-time safety monitoring—ensuring gear compliance and reducing industrial risks
Project Overview
This project focused on developing a long-range Personal Protective Equipment (PPE) compliance detection system using YOLOv5. The system was designed to automatically identify essential safety equipment such as hard hats and safety vests in industrial and construction environments. The goal was to improve safety compliance and reduce workplace hazards through real-time automated detection.
Problem Statement
Non-compliance with PPE regulations significantly increases the risk of injury in hazardous work environments. Existing monitoring solutions often fail to adequately ensure compliance, especially on large industrial or construction sites. There was a critical need for an automated, reliable system that could continuously monitor and ensure PPE compliance without constant human oversight.
Key Findings
- High-Accuracy Model Training: Successfully trained a YOLOv5 model to detect hard hats and safety vests across diverse environmental and lighting conditions, maintaining accuracy in real-world scenarios.
- Data Quality Was Crucial: Meticulous data collection and annotation significantly improved model accuracy, ensuring robust detection performance in varied site conditions.
- Real-Time Compliance Monitoring: Established optimal techniques for maintaining high detection rates in dynamic and cluttered environments, enabling real-time compliance verification.
Implemented Solution
Developed a custom dataset for hard hats and vests, used YOLOv5 for real-time compliance monitoring, and integrated an alert system to notify supervisors of non-compliance:
-
Custom PPE Dataset Development:
Created a dedicated dataset featuring annotated images of workers wearing or missing hard hats and safety vests, enabling targeted model training.
-
YOLOv5 Integration for Real-Time Detection:
Utilised YOLOv5’s object detection capabilities to ensure fast, accurate identification of PPE items across live video feeds.
-
Supervisor Alert System:
Implemented an automated alert system that notifies site supervisors instantly in cases of non-compliance, supporting proactive intervention and improved workplace safety.
Results
The AI-powered system achieved an 87% accuracy rate in identifying PPE compliance and non-compliance, providing timely alerts and visual logs to supervisors. As a result, incidents related to PPE non-compliance were reduced by 40%, significantly improving on-site safety. The system enabled proactive risk management and supported ongoing safety audits with data-driven insights. Its scalable design allowed seamless integration across various sites and environments, making it a valuable asset for improving worker protection and maintaining regulatory compliance.