• Laser & Optoelectronics Progress
  • Vol. 58, Issue 22, 2210008 (2021)
Binzhou Wang and Zhiyong Xiao*
Author Affiliations
  • School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, China
  • show less
    DOI: 10.3788/LOP202158.2210008 Cite this Article Set citation alerts
    Binzhou Wang, Zhiyong Xiao. Channel Attention Multi-Branch Network for Fine-Grained Image Recognition[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2210008 Copy Citation Text show less
    Structure of the RA-CNN
    Fig. 1. Structure of the RA-CNN
    Structure of our network
    Fig. 2. Structure of our network
    Structure of the ECA module
    Fig. 3. Structure of the ECA module
    Workflow and visualization results of the AAPM. (a) APN; (b) AAPM
    Fig. 4. Workflow and visualization results of the AAPM. (a) APN; (b) AAPM
    Example image of the datasets. (a) Food-101 dataset; (b) Stanford Cars dataset
    Fig. 5. Example image of the datasets. (a) Food-101 dataset; (b) Stanford Cars dataset
    Visualization of processing results. (a) Car 1; (b) car 2; (c) food 1; (d) food 2
    Fig. 6. Visualization of processing results. (a) Car 1; (b) car 2; (c) food 1; (d) food 2
    MethodStanford CarsFood-101
    Without ECA sub-network94.888.5
    Without DO-Conv95.289.8
    Without AAPM92.585.4
    Ours95.490.6
    Table 1. Ablation experimental results of our method on different datasets unit: %
    MethodAdded trainingAccuracy
    Bilinear-CNN(2015)×91.3
    RA-CNN(2017)×92.5
    HS-Net(2017)93.8
    AAA Model (2019)95.4
    MMAL(2020)×95.0
    Ours×95.4
    Table 2. Experimental results of different methods on the Stanford Cars dataset unit: %
    MethodAdded trainingAccuracy
    Bilinear-CNN(2015)×84.7
    WISeR(2018)×90.3
    FPCNN(2018)×87.9
    FCA(2019)86.3
    Ours×90.6
    Table 3. Experimental results of different methods on the Food-101 dataset unit: %
    Binzhou Wang, Zhiyong Xiao. Channel Attention Multi-Branch Network for Fine-Grained Image Recognition[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2210008
    Download Citation