Volume 17, Number 2
Research on Person and Vehicle Detection and Counting Method based on Improved YOLOV5
Authors
Zhidan Yuan, Jialu Sun , Yinglu Wei and Yuqing Zhang , Jiangsu Maritime Institute, China
Abstract
With the rapid advancement of autonomous driving technologies, Person and vehicle detection and counting tasks in urban road environments are confronted with significant challenges, including object occlusion, background interference, and insufficiently discriminative feature representations. These factors directly degrade detection accuracy and counting reliability. To address these issues, this study proposes an improved algorithm that integrates YOLOv5 with the SE attention mechanism. The SE module adaptively recalibrates channel-wise feature responses by modeling inter-channel dependencies, thereby enhancing the network’s focus on salient target features while suppressing irrelevant background noise and interference. Experiments were conducted on the BDD100K dataset to evaluate the effectiveness of the proposed approach. The results demonstrate that the improved YOLOv5 model outperforms the baseline YOLOv5 in Person and vehicle detection and counting tasks, achieving a performance improvement of 6.63%. These findings indicate that the proposed method enhances both detection accuracy and robustness, and thus exhibits greater efficiency and reliability for practical deployment in autonomous driving systems.
Keywords
Object Detection, YOLOv5, Attention Mechanism, Human and Vehicle Detection and Counting