• Infrared and Laser Engineering
  • Vol. 53, Issue 1, 20230472 (2024)
Shan Xue1,2, Hongyu An1, Qiongying Lv1, and Guohua Cao2
Author Affiliations
  • 1College of Mechanical and Electrical Engineering, Changchun University of Science and Technology, Changchun 130022, China
  • 2Chongqing Research Institute, Changchun University of Science and Technology, Chongqing 400000, China
  • show less
    DOI: 10.3788/IRLA20230472 Cite this Article
    Shan Xue, Hongyu An, Qiongying Lv, Guohua Cao. Image target detection algorithm based on YOLOv7-tiny in complex background[J]. Infrared and Laser Engineering, 2024, 53(1): 20230472 Copy Citation Text show less
    Partial picture of drone dataset
    Fig. 1. Partial picture of drone dataset
    DUT-ANTI-UAV dataset
    Fig. 2. DUT-ANTI-UAV dataset
    Network chart of YOLOv7-tiny
    Fig. 3. Network chart of YOLOv7-tiny
    Network chart of YOLOv7-drone
    Fig. 4. Network chart of YOLOv7-drone
    Multi-scale channel attentional mechanism module
    Fig. 5. Multi-scale channel attentional mechanism module
    RFB network structure diagram
    Fig. 6. RFB network structure diagram
    Deformable convolution diagram
    Fig. 7. Deformable convolution diagram
    Comparison chart of test results before and after adding attention mechanism
    Fig. 8. Comparison chart of test results before and after adding attention mechanism
    Before and after adding RFB structure
    Fig. 9. Before and after adding RFB structure
    Before and after the introduction of small target detection layer detection results map
    Fig. 10. Before and after the introduction of small target detection layer detection results map
    Before and after adding SIoU structure
    Fig. 11. Before and after adding SIoU structure
    Before and after adding DCN structure
    Fig. 12. Before and after adding DCN structure
    Comparison of detection performance of different algorithms
    Fig. 13. Comparison of detection performance of different algorithms
    Before and after the algorithm improvement CAM contrast chart
    Fig. 14. Before and after the algorithm improvement CAM contrast chart
    ModelParams/MmAP@0.5FPS/frame·s-1GFLOPS
    YOLOv7-tiny6.0284.37413.2
    +SE6.0584.57513.2
    +CBAM6.0284.97313.3
    +EMA6.0685.97513.5
    +SMSE8.9986.87115.6
    Table 1. Comparison of detection performance of different attention mechanism algorithms
    ModelParams/MmAP@0.5FPS/frame·s-1GFLOPS
    YOLOv7-tiny6.0284.37413.2
    +XMB6.3085.67714.7
    Table 2. The RFB structure algorithm is introduced to detect the performance comparison
    ModelParams/MmAP@0.5FPS/frame·s-1GFLOPS
    YOLOv7-tiny6.0284.37413.2
    +XMB6.1085.17015.5
    Table 3. The small target detection layer algorithm is introduced to detect the performance comparison
    ModelParams/MmAP@0.5FPS/frame·s-1GFLOPS
    YOLOv7-tiny6.0284.37413.2
    +SIoU6.0285.07813.2
    Table 4. Improved loss function algorithm detection performance comparison
    ModelParams/MmAP@0.5FPS/frame·s-1GFLOPS
    YOLOv7-tiny6.0284.37413.2
    +DCN6.0886.67114.8
    Table 5. A comparison of detection performance before and after deformable convolution is introduced
    SMSERFBXMBSIoUDCNParams/MmAP@0.5FPS/frame·s-1GFLOPS
    6.0284.37413.2
    8.9986.87115.6
    9.2987.47317.1
    9.3988.26920.1
    9.3988.77320.1
    9.4590.47221.7
    Table 6. Gradually add each module algorithm detection performance comparison
    ModelParams/MmAP@0.5FPS/frame·s-1GFLOPS
    YOLOv7-tiny6.0284.37413.2
    YOLOv737.287.357104.8
    YOLOv5l46.186.442107.9
    YOLOv7-drone9.4590.47221.7
    Table 7. Comparison of detection performance of different target detection algorithms
    ModelParams/MmAP@0.5FPS/frame·s-1GFLOPS
    YOLOv7-tiny6.0265.07413.2
    YOLOv7-drone9.4571.07221.7
    Table 8. PASCAL VOC dataset detection performance comparison
    Shan Xue, Hongyu An, Qiongying Lv, Guohua Cao. Image target detection algorithm based on YOLOv7-tiny in complex background[J]. Infrared and Laser Engineering, 2024, 53(1): 20230472
    Download Citation