Journal of Applied Optics, Volume. 45, Issue 2, 405(2024)

Few shot target detection method based on background suppression and classification correction

Wei CAI, Xin WANG*, Xinhao JIANG, Zhiyong YANG, and Dong CHEN
Author Affiliations
  • School of Missile Engineering, Rocket Military Engineering University, Xi'an 710025, China
  • show less
    Figures & Tables(12)
    Architecture diagram of few shot target detection methods for background suppression and classification correction
    Structure diagram of background suppression module
    Structure diagram of feature aggregation module
    Structure diagram of detection head introducing contrast branch
    Schematic diagram of comparison of three detection heads
    Comparison images of algorithm results of military datasets
    RQ4 target visualization images for each attitude
    Comparison images of algorithm results of PASCAL VOC datasets
    • Table 1. Hardware platform configuration

      View table
      View in Article

      Table 1. Hardware platform configuration

      名称相关配置
      GPUNVIDIA GV100
      CPU4210/128G
      GPU内存32G
      操作系统Win10
      计算平台CUDA10.2
      CPU(测试)Inter Core i7 10700/16G
    • Table 2. Experimental results of different target detection algorithms based on few shot dataset of military airborne incoming targets

      View table
      View in Article

      Table 2. Experimental results of different target detection algorithms based on few shot dataset of military airborne incoming targets

      方法/shotMAP@0.5
      1 shot2 shot3 shot5 shot10 shot
      TFA/fc25.629.427.437.144.2
      TFA/cos17.223.523.123.429.2
      Attention RPN23.829.229.139.454.3
      本文算法28.332.839.942.956.2
    • Table 3. Ablation experiment based on few shot dataset of military airborne incoming targets

      View table
      View in Article

      Table 3. Ablation experiment based on few shot dataset of military airborne incoming targets

      基线 模型背景 抑制特征 聚合对比 分支MAP@0.5
      1 shot2 shot3 shot5 shot10 shot
      25.629.427.437.144.2
      27.329.736.741.454.1
      27.530.137.642.554.4
      28.332.839.942.956.2
    • Table 4. Experimental results of different target detection algorithms based on PASCAL VOC dataset

      View table
      View in Article

      Table 4. Experimental results of different target detection algorithms based on PASCAL VOC dataset

      方法/shotMAP@0.5
      1 shot2 shot3 shot5 shot10 shot
      TFA/fc36.839.143.655.757.0
      TFA/cos39.836.144.755.756.0
      Attention RPN35.036.039.151.755.7
      本文算法40.147.648.057.160.5
    Tools

    Get Citation

    Copy Citation Text

    Wei CAI, Xin WANG, Xinhao JIANG, Zhiyong YANG, Dong CHEN. Few shot target detection method based on background suppression and classification correction[J]. Journal of Applied Optics, 2024, 45(2): 405

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Research Articles

    Received: Jun. 7, 2023

    Accepted: --

    Published Online: May. 28, 2024

    The Author Email: WANG Xin (王鑫(1999—))

    DOI:10.5768/JAO202445.0203002

    Topics