Pedestrian Target Detection Based on Attention Mechanism in Cloud Computing

As one of the most fundamental facial recognition technologies, pedestrian target detection faces particular challenges in UAV (unmanned aerial vehicle) aerial images where small targets are often shadowed or obscured. This paper proposes a Convolutional Block Attention Module (CBAM) UAV pedestrian detection resource allocation strategy for edge devices to tackle these challenges. In the study, the CBAM module was first added to enhance feature expression and improve the model's robustness. Second, the Scale Invariant Objective (SIOU) loss function was employed to raise recognition precision, resulting in a new generation of YOLOv8 pedestrian target detection based on the attention mechanism. Finally, extensive experiments were performed on the YOLOv8 baseline models, with results on our dataset demonstrating that the proposed strategy achieved state-of-the-art performance and efficiency across various model scales.