<![CDATA[Workplace safety remains a critical concern in high-risk industrial sectors due to frequent non-compliance with Personal Protective Equipment (PPE) regulations. This study aims to develop and evaluate a real-time, web-based system for automatically detecting compliance with two essential PPE items, safety helmets and safety vests, among workers in industrial work environments, including its detection accuracy, generalization on unseen data, and notification reliability. This research adopts a Research and Development (R&D) approach that combines a machine-learning experimentation pipeline for model development with a software prototyping process for the monitoring application, encompassing dataset collection from Roboflow (2,777 images across six classes: Hardhat, No Hardhat, Safety Vest, No Safety Vest, Person, and Background), data preprocessing, training of a YOLOv8n model (100 epochs, image size 800×800, batch size 16, SGD optimizer), performance evaluation on the validation set, qualitative functional testing, and implementation into a Flask-based website prototype integrated with Telegram Bot API notifications. On the validation set, the trained YOLOv8n model achieved a mean Average Precision at IoU 0.5 (mAP50) of 0.681 and an mAP50-95 of 0.416, at the best-fitness checkpoint (epoch 77 of 100). At the operational confidence threshold of 0.262 identified from the F1-Confidence curve, the model achieved an F1-Score of 0.69. Independent evaluation on the held-out 84-image test set is identified as a priority for follow-up work. Novelty: Unlike prior studies that generally stop at the model training and evaluation stage, and unlike [18], which implemented a YOLO-based PPE completeness dashboard without an automated alerting mechanism, this research integrates the YOLOv8 detection model directly into a real-time monitoring dashboard with an automatic Telegram text-and-image notification mechanism, enabling field supervisors to receive instant alerts whenever a PPE violation is detected. The resulting prototype demonstrates initial feasibility for supporting PPE compliance monitoring in industrial workplaces, pending more extensive field validation.]]>
Copyrights © 2026
