• Laser & Optoelectronics Progress
  • Vol. 57, Issue 6, 061008 (2020)
Zhong Ji, Qiankun Kong, and Jian Wang*
Author Affiliations
  • School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China
  • show less
    DOI: 10.3788/LOP57.061008 Cite this Article Set citation alerts
    Zhong Ji, Qiankun Kong, Jian Wang. Object Detection Algorithm Guided by Dual Attention Models[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061008 Copy Citation Text show less
    Architecture of the proposed DAGM model
    Fig. 1. Architecture of the proposed DAGM model
    Architecture of MFCA model
    Fig. 2. Architecture of MFCA model
    Visualization of the feature map
    Fig. 3. Visualization of the feature map
    Architecture of SCSA model
    Fig. 4. Architecture of SCSA model
    AP histogram of different detection algorithms
    Fig. 5. AP histogram of different detection algorithms
    MethodBackboneAPAP50AP75APSAPMAPL
    Faster R-CNN+++[8]ResNet-101-C434.955.737.415.638.750.9
    Faster R-CNN (FR) byG-RMI[28]ResNet-101-C434.755.536.713.538.152.0
    YOLOv2[24]DarkNet-1921.644.019.25.022.435.5
    SSD513[4]ResNet-101-SSD31.250.433.310.234.549.8
    DSSD513[26]ResNet-101-DSSD33.253.335.213.035.451.1
    RON[31]VGG-1627.427.149.5---
    DeNet[29]DeNet-10133.853.436.112.336.150.8
    CoupleNet[22]ResNet-10133.153.535.411.636.350.1
    YoLov3[25]DarkNet-5333.057.934.418.335.441.9
    SIN[23]VGG-1623.244.522.07.324.536.3
    Relation Network[32]ResNet-5032.554.033.8---
    MLKP[30]ResNet-10126.948.426.98.629.241.1
    MFCA (ours)ResNet-10136.254.538.718.539.247.6
    Table 1. Comparison of detection results with different detection algorithms%
    ModuleBackboneAPAP50AP75APSAPMAPL
    BaselineResNet-10135.052.637.717.138.848.0
    Baseline+MFCAResNet-10135.553.338.118.439.248.1
    Baseline+MFCA+SCSAResNet-10135.954.038.618.839.848.5
    Table 2. Comparison of test results of each module%
    Zhong Ji, Qiankun Kong, Jian Wang. Object Detection Algorithm Guided by Dual Attention Models[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061008
    Download Citation