ASPPF-YOLO: A Pedestrian and Vehicle Detection Algorithm Based on Attention Mechanism

Pedestrian and vehicle detection is a key issue in intelligent transportation, smart logistics, autonomous driving and other fields. However, existing methods either require high device performance and long rounds of model training, or have limited accuracy and robustness. To address this issue, we propose ASPPF-YOLO, a detection algorithm based on the attention mechanism. Firstly, we improve SPPF by using Convolutional Block Attention Module (CBAM). In addition, Wise-IoU (WIOU) loss function is introduced to fully unleash the potential of the non-monotonic focusing mechanism. Experimental evaluations were conducted on foggy and general road datasets. The results show that ASPPF-YOLO algorithm exhibits superior performance, verifying its effectiveness and reliability in detecting vehicles and pedestrians under the condition of small-scale training costs.