Advancing Industrial Safety: A Spatio-Temporal Framework for PPE Detection Using YOLOv11

Ensuring adherence to personal protective equipment (PPE) regulations is paramount for workplace safety in industrial settings. This study presents an innovative framework that integrates YOLOv11 with the spatiotemporal compliance algorithm (STCA) to enhance PPE monitoring by combining advanced spatial detection with temporal reasoning. A dataset of $\mathbf{5 0, 0 0 0}$ annotated images and video frames from various industrial environments was utilized for training and evaluation. Key innovations include the temporal entropy minimization function (TEMF), Adaptive compliance scoring function (ACSF), and dynamic PPE compliance loss (DPCL), which collectively improve detection accuracy and temporal consistency. Experimental results indicate that YOLOv11+STCA achieves a compliance accuracy of $\mathbf{9 9. 2 \%}$, significantly outperforming existing models such as YOLOv8 and standalone YOLOv11. This framework marks a substantial advancement in industrial safety, providing a robust and scalable solution for real-time monitoring in dynamic environments. It addresses critical shortcomings in existing systems, paving the way for enhanced workplace safety and broader applications in other safety-critical domains.