• Infrared and Laser Engineering
  • Vol. 52, Issue 8, 20230245 (2023)
Xuezhi Zhang1, Hongdong Zhao1,2, Weina Liu1, Yiming Zhao1, and Song Guan2
Author Affiliations
  • 1School of Electronic and Information Engineering, Hebei University of Technology, Tianjin 300401, China
  • 2National Key Laboratory of Electromagnetic Space Security, Tianjin 300308, China
  • show less
    DOI: 10.3788/IRLA20230245 Cite this Article
    Xuezhi Zhang, Hongdong Zhao, Weina Liu, Yiming Zhao, Song Guan. An infrared vehicle detection method based on improved YOLOv5[J]. Infrared and Laser Engineering, 2023, 52(8): 20230245 Copy Citation Text show less
    Improved YOLOv5 network architecture diagram
    Fig. 1. Improved YOLOv5 network architecture diagram
    Structure of CFG
    Fig. 2. Structure of CFG
    Structure of CA
    Fig. 3. Structure of CA
    Structure of Spatial Attention
    Fig. 4. Structure of Spatial Attention
    (a) FPN structure, adds an upward path from small-sized feature map; (b) PANet structure, adds a downward path from large-sized feature map based on FPN; (c) BiFPN structure; (d) Z-BiFPN
    Fig. 5. (a) FPN structure, adds an upward path from small-sized feature map; (b) PANet structure, adds a downward path from large-sized feature map based on FPN; (c) BiFPN structure; (d) Z-BiFPN
    Diagram of Decoupled Head architecture
    Fig. 6. Diagram of Decoupled Head architecture
    (a) An example of a self-collected dataset; (b) An example of SCUT_FIR_Pedestrian_Dataset; (c) An example of MULTISPECTRAL DATASET
    Fig. 7. (a) An example of a self-collected dataset; (b) An example of SCUT_FIR_Pedestrian_Dataset; (c) An example of MULTISPECTRAL DATASET
    PR curve. (a) YOLOv5; (b) Improved YOLOv5
    Fig. 8. PR curve. (a) YOLOv5; (b) Improved YOLOv5
    Confusion matrices. (a) YOLOv5; (b) Improved YOLOv5
    Fig. 9. Confusion matrices. (a) YOLOv5; (b) Improved YOLOv5
    Comparison of detection results. (a) Original image; (b) YOLOv5; (c) Improved YOLOv5
    Fig. 10. Comparison of detection results. (a) Original image; (b) YOLOv5; (c) Improved YOLOv5
    abcde
    YOLOv5s
    CFG
    Four Head
    Z-BiFPN
    Decoupled Head
    AP-bus88.5%85.3%87.1%86.0%89.4%
    AP-truck81.0%81.2%82.9%81.0%85.4%
    AP-car89.6%88.7%89.1%89.6%90.3%
    AP-van78.3%76.4%77.8%79.8%82.6%
    AP-person79.3%82.6%81.0%79.7%83.5%
    AP-bicycle72.0%76.6%75.7%79.5%86.2%
    AP-elecmot79.0%80.1%80.2%82.5%79.7%
    P86.5%89.4%85.8%86.9%88.2%
    R73.8%75.4%74.7%76.7%77.4%
    mAP81.1% 81.6% 82.0% 82.6% 85.3%
    Table 1. Experimental results of different improvement methods
    ModelsSSDYOLOv3YOLOv5YOLOR-W6YOLOv7-tinyYOLOXOurs
    AP-bus76.4%85.9%88.5%81.7%85.7%87.6%89.4%
    AP-truck88.0%83.8%81.0%82.3%82.4%84.2%85.4%
    AP-car68.7%83.3%89.6%90.1%90.8%90.3%90.3%
    AP-van63.2%71.9%78.3%80.3%79.2%82.1%82.6%
    AP-person35.8%70.1%79.3%76.9%75.1%81.5%83.5%
    AP-bicycle41.9%50.2%72.0%44.7%53.0%78.3%86.2%
    AP-elecmot47.3%65.6%79.0%80.5%65.3%80.7%79.7%
    mAP60.2%73.0%81.1%76.6%75.9%83.5%85.3%
    Parameters24.4×10661.6×1067.0×10679.3×1066.0×1068.9×10610.4×106
    Weight/MB93.7235.213.7151.811.717.320.3
    Table 2. Comparison of different object detection algorithms
    Xuezhi Zhang, Hongdong Zhao, Weina Liu, Yiming Zhao, Song Guan. An infrared vehicle detection method based on improved YOLOv5[J]. Infrared and Laser Engineering, 2023, 52(8): 20230245
    Download Citation