With the widespread investment in intelligent traffic management, vehicle-related object detection and tracking are the most needed step in computer research. However, complex road conditions result in dense vehicles and small intervals between vehicles on the road, it is hard to identify and detect vehicle types quickly and accurately. To overcome this issue, this paper proposes a YOLOX object detection algorithm to be applied to traffic road vehicle detection. First, the CSPnet and Focus network structures are added to the Darknet network to improve target detection accuracy. Secondly, the decoupling head and SimOTA strategy are added to the prediction end to reduce the training time and further enhance the target detection efficiency. The pre-trained weights on the COCO dataset are fused together perform detection and identification to enhance the detection generalization and accuracy. The results of experiments demonstrate the detection of vehicles on the bus dataset constructed in this paper, vehicle detection of multiple targets and small targets in dense vehicle scenarios.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    YOLOX-based dynamic vehicle target detection


    Contributors:


    Publication date :

    2022-10-12


    Size :

    2080227 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English




    Maritime Object Detection based on YOLOx for Aviation Image

    Wang, Yuan / Yuan, Haiwen / Li, Yongshuai et al. | IEEE | 2022


    Airport Runway Foreign Object Debris (FOD) Detection Based on YOLOX Architecture

    Taupik, Jajang / Alamsyah, Tossin / Wulandari, Asri et al. | IEEE | 2023


    Ship detection in infrared images via bounding boxes based on improved YOLOX

    Qiu, Xinjie / Li, Zhiyu / Han, Fenglei et al. | SPIE | 2022


    A Novel UAV Visual Positioning Algorithm Based on A-YOLOX

    Ying Xu / Dongsheng Zhong / Jianhong Zhou et al. | DOAJ | 2022

    Free access