Automated Personal Protective Equipment (PPE) Detection in Construction Environments: Enhancing Safety through YOLOv8 Object Detection

This research addresses the critical issue of Personal Protective Equipment (PPE) compliance in construction sites by developing an automated detection system. The study aims to improve worker safety through real-time detection of PPE, including hard hats, masks, overall suits, safety boots, and gloves. Using the YOLOv8 object detection model, trained on a custom dataset comprising images captured by the researchers and sourced from Roboflow, the system was tested under various environmental conditions. The model achieved its best performance at epoch 77 with a precision of 0.84976, a recall of 0.66429, a mean Average Precision (mAP) at 50% IoU (mAP50) of 0.72938, a mAP50-95 of 0.49907, and an F1 score of approximately 0.745. These metrics indicate substantial improvements in detection accuracy and balanced performance in both precision and recall. Future work will focus on enhancing the model's robustness to weather variations, addressing issues such as dirty clothing, low lighting, and motion blur to further improve detection accuracy and reliability.