• Laser & Optoelectronics Progress
  • Vol. 58, Issue 20, 2010002 (2021)
Kaixuan Wang1、*, Guanghua Chen1、2, and Hongjia Chu1
Author Affiliations
  • 1Microelectronics R&D Center, Shanghai University, Shanghai 200444, China;
  • 2School of Mechatronic Engineering and Automation, Shanghai University, Shanghai 200444, China
  • show less
    DOI: 10.3788/LOP202158.2010002 Cite this Article Set citation alerts
    Kaixuan Wang, Guanghua Chen, Hongjia Chu. Finger Vein Recognition Based on Improved ResNet[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2010002 Copy Citation Text show less
    Residual block
    Fig. 1. Residual block
    Conventional convolution
    Fig. 2. Conventional convolution
    Depthwise convolution
    Fig. 3. Depthwise convolution
    Depthwise over-parameterized convolution
    Fig. 4. Depthwise over-parameterized convolution
    Structure of dual attention mechanism
    Fig. 5. Structure of dual attention mechanism
    Improved residual block
    Fig. 6. Improved residual block
    Improved ResNet
    Fig. 7. Improved ResNet
    Test accuracy. (a) FV-USM;(b) SDUMLA
    Fig. 8. Test accuracy. (a) FV-USM;(b) SDUMLA
    MethodAccuracy /%Time /msParameter
    FV-USMSDUMLAFV-USMSDUMLA
    VGG-1695.833395.7721494944.0×106
    DenseNet96.951298.161834348.5×106
    AlexNet92.276496.3235161626.7×106
    ResNet-1896.036697.7941161619.7×106
    Improved ResNet+DO-Conv98.475698.7132151514.8×106
    Improved ResNet+DO-Conv+LSCE99.085498.8971151514.8×106
    Improved ResNet+DO-Conv+LSCE+(SE+SAM)99.491999.4485151514.8×106
    Table 1. Comparison of structure and performance indicators of different models
    Kaixuan Wang, Guanghua Chen, Hongjia Chu. Finger Vein Recognition Based on Improved ResNet[J]. Laser & Optoelectronics Progress, 2021, 58(20): 2010002
    Download Citation