YOLOv10n-SD:a novel real-time object detectionmodel for Driver distracted driving detection

Distracted driving is one of the leading causes of traffic accidents, and monitor-ing driver’s distracted states is crucial for improving road safety. In this paper,we adopt the Swin Transformer as the backbone network for YOLOv10. TheTransformer architecture effectively captures global dependencies between var-ious positions in the sequence, enhancing the model’s ability to capture globalobject relationships. Additionally, we design and implement a novel attentionmechanism module, DECS (Directional Enhanced Channel Spatial AttentionModule), to replace the SPA module in YOLOv10, which further strengthens themodel’s capability to identify critical features. We constructed a large-scale anddiverse driver monitoring image dataset,CBTDDD, which encompasses variousvehicle types, including buses, trucks, and sedans. Experimental results demon-strate that our model achieves a significant improvement of 4.3% in the mAP50metric, validating the effectiveness of integrating the Swin Transformer with theDECS module. This work provides a new technical pathway for distracted driv-ing detection, enhances detection accuracy and robustness, and holds substantialpractical application value.