Small Target Detection Based on YOLOX

DOI
  • Ren Keying
    School of Electronic Information and Automation, Tianjin University of Science and Technology
  • Chen Xiaoyan
    School of Electronic Information and Automation, Tianjin University of Science and Technology
  • Chen Zhihui
    School of Electronic Information and Automation, Tianjin University of Science and Technology

抄録

With the development of drone, small target detection has become a hotspot of current research. In this paper, the network of small target detection is based on YOLOX is studied. There are a lot of small targets in the images taken by UAV, which brings great difficulty to the detection task. The main improvement of the proposed network is that, based on the original YOLOX network, four-scale adaptive spatial feature fusion pyramid is added to filter the conflicting information between different scales and improve the expressive accuracy of the small target features. Experiments show that the proposed method has average accuracy of 32.83% in the self-built dataset, which is 2.53% higher than that of YOLOX-s, 1.53% higher than that of Yolov6-s, and 5.73% higher than that of YOLOv5-s.

収録刊行物

詳細情報 詳細情報について

  • CRID
    1390015333257550720
  • DOI
    10.5954/icarob.2023.os31-1
  • ISSN
    21887829
  • 本文言語コード
    en
  • データソース種別
    • JaLC
    • Crossref
  • 抄録ライセンスフラグ
    使用不可

問題の指摘

ページトップへ