SAFETYHELMETANDMASKDETECTIONALGORITHMBASEDONYOLOX

Deep learning- based algorithms for detecting safety helmets and masks often suffer from low accuracy and large model sizes. To address these limitations, this paper proposes a method for detecting safety helmets and masks based on YOLOX- S. This method combined a global contextual block (GCblock) and a feature enhancement module (RFB) based on the YOLOX- S algorithm. It incorporated shallow effective feature layers and attention modules into the enhanced feature extraction network, and replaced the Mish activation function. These improvements enhanced the model's information extraction ability. Experimental results show that the modified network model achieves the mAP value of 87.18%, which is a 1.07 percentage points improvement over the original YOLOX- S network model.