About | Contact Us | Register | Login
ProceedingsSeriesJournalsSearchEAI
Proceedings of the 2nd International Conference on Machine Learning and Automation, CONF-MLA 2024, November 21, 2024, Adana, Turkey

Research Article

Research on Improvement of Environment Perception Algorithm for Autonomous Driving Vehicles Based on YOLOv5

Download224 downloads
Cite
BibTeX Plain Text
  • @INPROCEEDINGS{10.4108/eai.21-11-2024.2354630,
        author={Yuxi  Yang},
        title={Research on Improvement of Environment Perception Algorithm for Autonomous Driving Vehicles Based on YOLOv5},
        proceedings={Proceedings of the 2nd International Conference on Machine Learning and Automation, CONF-MLA 2024, November 21, 2024, Adana, Turkey},
        publisher={EAI},
        proceedings_a={CONF-MLA},
        year={2025},
        month={3},
        keywords={object detection autonomous driving yolov5s lightweight convolution multi scale detection loss function},
        doi={10.4108/eai.21-11-2024.2354630}
    }
    
  • Yuxi Yang
    Year: 2025
    Research on Improvement of Environment Perception Algorithm for Autonomous Driving Vehicles Based on YOLOv5
    CONF-MLA
    EAI
    DOI: 10.4108/eai.21-11-2024.2354630
Yuxi Yang1,*
  • 1: Tongji University, Shanghai, China
*Contact email: yang3611@tongji.edu.cn

Abstract

Autonomous driving relies heavily on vehicle object detection, and YOLOv5s is presently one of the best algorithms for this purpose. However, in extreme environments such as severe weather, cars have poor perception of the environment, and their ability to detect dynamic targets is greatly affected, resulting in low accuracy and poor robustness of YOLOv5 object detection algorithm in pedestrian and vehicle detection. This article proposes an improved YOLOv5s algorithm. Firstly, a selective attention mechanism (SimAM) module is used to weight the output of the convolutional layer, allowing the network to quickly capture regions of interest and suppress irrelevant information; Simultaneously using lightweight convolution GSConv instead of the conventional convolution to compensate for semantic information loss and reduce model complexity; Secondly, adding a shallow detection layer changes the original algorithm's three scale detection to four scale detection, enhancing the learning ability for small-scale targets; Finally, SIoU Loss is used as the bounding box regression loss function to achieve more accurate localization of the predicted boxes. The improved YOLOv5s algorithm was tested on the CARLA simulation dataset, and simulation results showed that the average detection accuracy of the improved model reached 96.67%, which improved the detection accuracy for complex scenes.

Keywords
object detection autonomous driving yolov5s lightweight convolution multi scale detection loss function
Published
2025-03-11
Publisher
EAI
http://dx.doi.org/10.4108/eai.21-11-2024.2354630
Copyright © 2024–2025 EAI
EBSCOProQuestDBLPDOAJPortico
EAI Logo

About EAI

  • Who We Are
  • Leadership
  • Research Areas
  • Partners
  • Media Center

Community

  • Membership
  • Conference
  • Recognition
  • Sponsor Us

Publish with EAI

  • Publishing
  • Journals
  • Proceedings
  • Books
  • EUDL