Performance Analysis of Real-Time Safety Vest and Helmet Detection Using Yolov8

Workplace safety is a critical priority in the industry, particularly in enforcing occupational health and safety (OHS) standards. This study introduces an innovative OHS compliance inspection system integrating access control with object recognition technology. The system captures images of workers using an ESP32Cam as the pre-processing step for safety helmet and vest detection. The ESP32 Cam captures the best image for detection at a maximum distance of 100 cm. The OpenCV library accesses these images, while the YOLOv8 library performs object recognition. The dataset includes 761 images for helmets and 314 images for vests, with the best results achieved using 100 training epochs of the pre-trained yolov8s.pt model. The system demonstrates a confidence level of 0.89 and an accuracy rate 0.75 in identifying safety helmets and vests. Recognition results are monitored in real-time via a web-based dashboard, with data stored in a database for further analysis. This study highlights the potential of integrating object recognition technology to enhance OHS compliance and workplace safety.