An Efficient YOLOv8-Based Model With Cross-Level Path Aggregation Enabling Personal Protective Equipment Detection

Personal protective equipment (PPE) detection is an effective means to reduce potential safety hazards in intelligence monitoring for disaster prevention and security management. Addressing the limitations of existing methods, manual observation and inspection in the high-risk working environment of coal mines are challenging and time-consuming. Thus, a novel deep learning model based on YOLOv8 is proposed for PPE detection. First, an omnidimensional dynamic convolution is introduced into the backbone network, to avoid the generation of redundant features, thereby reducing network parameters and computational complexity. Meanwhile, a simple, parameter-free attention module is added to enhance the model's focus on critical features of equipment. Second, a bottom-up cross-level path aggregation is incorporated into the feature fusion structure of the Neck to minimize the loss of feature information for small targets. Finally, a novel convolutional connectivity integrated with depthwise separable convolution is applied in the Head to balance the computational burden brought by the improved modules. Experimental results demonstrate that the proposed model outperforms other state-of-the-art approaches: achieving a mean average precision of 92.68%, and reducing the parameter quantity to 9.09M. The proposed approach achieves a desired tradeoff between computational speed and recognition accuracy for PPE detection, providing robust support for coal mining safety production.