• Laser & Optoelectronics Progress
  • Vol. 58, Issue 8, 0810019 (2021)
Chengshuo Cao and Jie Yuan*
Author Affiliations
  • School of Electrical Engineering, Xinjiang University, Urumchi, Xinjiang 830047, China
  • show less
    DOI: 10.3788/LOP202158.0810019 Cite this Article Set citation alerts
    Chengshuo Cao, Jie Yuan. Mask-Wearing Detection Method Based on YOLO-Mask[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810019 Copy Citation Text show less
    YOLOv3 network structure diagram
    Fig. 1. YOLOv3 network structure diagram
    Residual unit and SE-residual unit
    Fig. 2. Residual unit and SE-residual unit
    Improved feature fusion network structure. (a) FPN; (b) bottom-up path aggregation
    Fig. 3. Improved feature fusion network structure. (a) FPN; (b) bottom-up path aggregation
    YOLO-Mask algorithm network structure
    Fig. 4. YOLO-Mask algorithm network structure
    Example of dataset images
    Fig. 5. Example of dataset images
    Comparison of detection results of different algorithms. (a)(b)(c) SSD detection results; (d)(e)(f) Faster R-CNN detection results; (g)(h)(i) YOLOv3 detection results; (j)(k)(l) YOLO-Mask detection results
    Fig. 6. Comparison of detection results of different algorithms. (a)(b)(c) SSD detection results; (d)(e)(f) Faster R-CNN detection results; (g)(h)(i) YOLOv3 detection results; (j)(k)(l) YOLO-Mask detection results
    Single target P-R curves of YOLO-Mask algorithm. (a) Right_mask; (b) wrong_mask; (c) no_mask
    Fig. 7. Single target P-R curves of YOLO-Mask algorithm. (a) Right_mask; (b) wrong_mask; (c) no_mask
    Batch sizeFactorPatienceLearn rate
    40.520.001
    Table 1. Selection of key parameters
    MethodAPmAP
    Right_maskWrong_maskNo_mask
    SSD67.3373.3465.1768.61
    Faster R-CNN75.7871.2772.9473.33
    YOLOv389.8179.9787.3485.71
    YOLO-Mask94.3091.1394.5693.33
    Table 2. Comparison of performance indicators of different detection algorithms unit: %
    Chengshuo Cao, Jie Yuan. Mask-Wearing Detection Method Based on YOLO-Mask[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810019
    Download Citation