• Laser & Optoelectronics Progress
  • Vol. 59, Issue 10, 1010005 (2022)
Zhiyong Song and Haipeng Pan*
Author Affiliations
  • School of Mechanical and Automatic Control, Zhejiang Sci-Tech University, Hangzhou 310018, Zhejiang , China
  • show less
    DOI: 10.3788/LOP202259.1010005 Cite this Article Set citation alerts
    Zhiyong Song, Haipeng Pan. Fabric Defect Classification Algorithm Based on Multi-Scale Feature Fusion of Spatial Attention[J]. Laser & Optoelectronics Progress, 2022, 59(10): 1010005 Copy Citation Text show less
    Overall network structure
    Fig. 1. Overall network structure
    Fabric defect types of printed fabric data set
    Fig. 2. Fabric defect types of printed fabric data set
    Training process of printed fabric. (a) Accuracy curve; (b) loss curve
    Fig. 3. Training process of printed fabric. (a) Accuracy curve; (b) loss curve
    Confusion matrix
    Fig. 4. Confusion matrix
    Positioning effect of printed fabric. (a) Original picture; (b) heatmap; (c) location result
    Fig. 5. Positioning effect of printed fabric. (a) Original picture; (b) heatmap; (c) location result
    ParameterValue
    Learning rate0.0001
    Batch size5
    Epoch50
    Image dimention256×256×1
    Table 1. Super parameter setting
    Defect typePrecision /%Recall /%Specificity /%
    Average93.2392.598.5
    Normal82.6195.0096.00
    Hole9595.0099.00
    Stain10095.00100.00
    Scratch100100.00100.00
    Thread ends81.8290.0096.00
    Fold10080.00100.00
    Table 2. Defect detection index of each type of printed fabric
    AlgorithmPrecision /%Recall /%Specificity /%Accuracy /%
    VGG1657.3356.6791.3356.67
    ResNet5049.93659365
    Proposed algorithm93.2392.598.592.5
    Table 3. Result comparison between proposed algorithm and other algorithms
    Zhiyong Song, Haipeng Pan. Fabric Defect Classification Algorithm Based on Multi-Scale Feature Fusion of Spatial Attention[J]. Laser & Optoelectronics Progress, 2022, 59(10): 1010005
    Download Citation