Research on Vehicle and Pedestrian Detection Algorithms Based on an Improved YOLOv8

DOI: https://doi.org/10.62517/jike.202504424

Author(s)

Yiming Sun

Affiliation(s)

School of Integrated Circuits, Nanjing University of Information Science & Technology, Suzhou, Jiangsu, China

Abstract

Vehicle and pedestrian detection in complex road scenarios represents a critical challenge for autonomous driving environmental perception. Addressing issues such as reduced model reliability caused by interference from lighting variations, occlusions, and adverse weather conditions. This paper addresses this critical issue by employing the YOLOv8 network as the foundational detection architecture, incorporating an attention mechanism to enhance the model's ability to extract, learn, and represent key object features. Based on this dataset, the experimental section presents two comparison approaches: one directly employs the original YOLOv8 model, while the second integrates the SE (Squeeze-and-Excitation) attention mechanism into the YOLOv8 model. Experimental results demonstrate that incorporating the SE attention mechanism achieves improvements in key metrics such as recall and mAP50-95, at the modest cost of increasing model parameters by 0.1 million and reducing inference speed by 3.4 FPS. Consequently, the proposed model exhibits superior overall performance and application potential in scenarios demanding high detection accuracy and real-time capabilities.

Keywords

YOLOV8; SE; Vehicle and Pedestrian Detection

References

[1] Tian Xuan, Wang Liang, Ding Qi. A Review of Deep Learning-Based Image Semantic Segmentation Methods [J]. Journal of Software, 2019, 30(02): 440-468. [2] Shao, Y. H., Zhang, D., Chu, H. Y., et al. Review of Deep Learning-Based YOLO Object Detection [J]. Transactions of Electronics and Information Technology, 2022, 44(10): 3697-3708. [3] Zhao Yongqiang, Rao Yuan, Dong Shipeng, et al. Review of Deep Learning Object Detection Methods [J]. Chinese Journal of Image and Graphics, 2020, 25(04): 629-654. [4] Chen Xianchang. Research on Deep Learning Algorithms and Applications Based on Convolutional Neural Networks [D]. Zhejiang Gongshang University, 2014. [5] Lu Hongtao, Zhang Qinchuan. A Review of Deep Convolutional Neural Networks in Computer Vision Applications [J]. Data Acquisition and Processing, 2016, 31(01): 1-17. [6] Li Bingzhen, Liu Ke, Gu Jiaojiao, et al. Research Review on Convolutional Neural Networks [J]. Computer Era, 2021, (04): 8-12+17. [7] Zhang Ke, Feng Xiaohan, Guo Yurong, et al. Review of Deep Convolutional Neural Network Models for Image Classification [J]. Chinese Journal of Image and Graphics, 2021, 26(10): 2305-2325. [8]Ren Huan, Wang Xuguang. Review of Attention Mechanisms [J]. Computer Applications, 2021, 41(S1): 1-6. [9] Han Qiang. Research on an Improved YOLOv8 Algorithm for Small Object Detection [D]. Jilin University, 2023.