• Laser & Optoelectronics Progress
  • Vol. 57, Issue 10, 101014 (2020)
Jian Xu, Shupei Wu*, and Xiuping Liu
Author Affiliations
  • School of Electronics Information, Xi'an Polytechnic University, Xi'an, Shaanxi 710048, China
  • show less
    DOI: 10.3788/LOP57.101014 Cite this Article Set citation alerts
    Jian Xu, Shupei Wu, Xiuping Liu. Classification of Bobbins Based on Improved Deep Neural Network[J]. Laser & Optoelectronics Progress, 2020, 57(10): 101014 Copy Citation Text show less
    AlexNet model structure
    Fig. 1. AlexNet model structure
    Activation function
    Fig. 2. Activation function
    Dataset samples. (a)(b) Black; (c)(d) yellow; (e)(f) green; (g)(h) brown; (i)(j) blackish green; (k)(l) orange; (m)(n) blue; (o)(p) purple
    Fig. 3. Dataset samples. (a)(b) Black; (c)(d) yellow; (e)(f) green; (g)(h) brown; (i)(j) blackish green; (k)(l) orange; (m)(n) blue; (o)(p) purple
    Neural network training process
    Fig. 4. Neural network training process
    Improved neural network model structure
    Fig. 5. Improved neural network model structure
    Impact of training sample number on test accuracy
    Fig. 6. Impact of training sample number on test accuracy
    Convolution feature map. AlexNet model (a) layer 1 feature map, (b) layer 3 feature map, (c) layer 5 feature map; improved model (d) layer 1 feature map, (e) layer 3 feature map, (f) layer 5 feature map
    Fig. 7. Convolution feature map. AlexNet model (a) layer 1 feature map, (b) layer 3 feature map, (c) layer 5 feature map; improved model (d) layer 1 feature map, (e) layer 3 feature map, (f) layer 5 feature map
    Accuracy and loss curves of AlexNet model. (a) Training and test accuracy; (b) loss curve
    Fig. 8. Accuracy and loss curves of AlexNet model. (a) Training and test accuracy; (b) loss curve
    Accuracy and loss curves of improved model. (a) Training and test accuracy; (b) loss curve
    Fig. 9. Accuracy and loss curves of improved model. (a) Training and test accuracy; (b) loss curve
    NetworkLayerConv1Pooling1Conv2Conv3Pooling2Conv4Conv5Conv6Pooling3
    I83×33×33×33×33×33×33×33×3
    II85×53×35×55×53×35×55×53×3
    III83×32×23×33×32×23×33×32×2
    IV93×33×33×33×33×33×33×33×33×3
    Table 1. Four kinds of network structures based on AlexNet
    NetworkIIIIIIIV
    Training time /h2.42.12.73.1
    Accuracy rate /%89.3181.7485.2088.42
    Table 2. Four kinds of network performance
    LayerLayerinputConvolution kernelConvolutionoutputPoolingLayer output
    SizeNumberStepSizeStep
    Conv1127×127×33×3641127×127×64127×127×64
    Pooling1127×127×643×3263×63×64
    Conv263×63×643×3128163×63×12863×63×128
    Conv363×63×1283×3128163×63×12863×63×128
    Pooling263×63×1283×3231×31×128
    Conv431×31×1283×3256131×31×25631×31×256
    Conv531×31×2563×3256131×31×25631×31×256
    Pooling331×31×2563×3215×15×256
    FC115×15×256256
    FC22568
    Table 3. Improved AlexNet model parameters
    ModelAlexNetImproved model without techniqueImproved model with technique
    Amount of error564328
    Error rate0.3730.2860.186
    Table 4. Impact of optimization techniques on model
    AlgorithmTraining time /hTraining accuracy /%Test accuracy /%Number of parameters /M
    Algorithm of Ref. [4]0.281.674.3
    Algorithm of Ref. [7]3.395.382.625.2
    AlexNet5.498.161.260.3
    VGG-166.399.267.7138.1
    Improved algorithm2.491.588.215.8
    Table 5. Comparison of recognition performance of different methods
    Jian Xu, Shupei Wu, Xiuping Liu. Classification of Bobbins Based on Improved Deep Neural Network[J]. Laser & Optoelectronics Progress, 2020, 57(10): 101014
    Download Citation