DBS-YOLO: A vehicle detection model based on improved YOLOv8 for UAV aerial scenes

Unmanned aerial vehicle (UAV) aerial images provide strong data support for vehicle detection tasks. However, due to diverse shooting angles, a high proportion of small targets, and susceptibility to interference from complex environments, most vehicle detection models in UAV aerial scenes fail to meet practical task requirements. Therefore, this study optimizes the YOLOv8 model and constructs DBS-YOLO. We propose the BiP2 detection layer to improve the detection sensitivity of small targets, employ SE Attention to focus on specific channel information and reduce interference from complex backgrounds, and utilize the C2f_DCN module to effectively handle irregularly shaped detection targets and reduce occurrences of false positives and false negatives. The experimental results show that our model demonstrates excellent detection ability by reducing the number of parameters while improving precision by $2.6 \%$, recall by $3.6 \%$, mAP@0.5 by 4.6%, and mAP@0.5:0.95 by 3.4%.