• Laser & Optoelectronics Progress
  • Vol. 59, Issue 16, 1615002 (2022)
Yuqing Liu1、2, Yaru Wang1、2、*, and Luyao Huang1、2
Author Affiliations
  • 1College of Engineering Science and Technology, Shanghai Ocean University, Shanghai 201306, China
  • 2Shanghai Engineering Research Center of Marine Renewable Energy, Shanghai 201306, China
  • show less
    DOI: 10.3788/LOP202259.1615002 Cite this Article Set citation alerts
    Yuqing Liu, Yaru Wang, Luyao Huang. Fish Recognition and Detection Based on FML-Centernet Algorithm[J]. Laser & Optoelectronics Progress, 2022, 59(16): 1615002 Copy Citation Text show less
    Network model structure of proposed algorithm
    Fig. 1. Network model structure of proposed algorithm
    Network structure of FFM
    Fig. 2. Network structure of FFM
    Image space geometric transformation. (a) Original image; (b) mirror image; (c) flip; (d) spin
    Fig. 3. Image space geometric transformation. (a) Original image; (b) mirror image; (c) flip; (d) spin
    Label data
    Fig. 4. Label data
    Target dataset
    Fig. 5. Target dataset
    Loss function curve of Centernet
    Fig. 6. Loss function curve of Centernet
    Loss function curve of FMCenternet
    Fig. 7. Loss function curve of FMCenternet
    Loss function curve of FML-Centernet
    Fig. 8. Loss function curve of FML-Centernet
    Partial recognition results. (a) (b) Faster-RCNN; (c) (d) YOLOV4; (e) (f) FML-Centernet
    Fig. 9. Partial recognition results. (a) (b) Faster-RCNN; (c) (d) YOLOV4; (e) (f) FML-Centernet
    AlgorithmAP75AP50AP
    Centernet28.6579.2138.03
    FMCenternet31.5181.3637.98
    FML-Centernet42.5185.9943.91
    Table 1. Comparative analysis experiments of different algorithms
    Network modelNetwork structureAP50 /%Average detection time /ms
    Faster-RCNNVGG74.32221.35
    YOLOV4CSPDarknet5376.95185.46
    CenternetResNet5079.21178.69
    FMCenternetResNet50+FFM81.36140.23
    FML-CenternetResNet50+FFM+LOSS85.9999.85
    Table 2. Comparative analysis of experimental results of different network models
    Network modelFishAP50 %Average detection time /ms
    Faster-RCNNSPT73.82220.36
    Pomfret74.21218.23
    Linnaeus73.34219.34
    CT74.53215.26
    DF74.15217.35
    YOLOV4SPT76.96180.52
    Pomfret76.52181026
    Linnaeus76.03184.37
    CT77.01183.45
    DF76.86187.25
    FML-CenternetSPT85.45100.21
    Pomfret86.1398.36
    Linnaeus86.0298.45
    CT85.8696.15
    DF86.2497.69
    Table 3. Verification results of different network models on target test set
    Yuqing Liu, Yaru Wang, Luyao Huang. Fish Recognition and Detection Based on FML-Centernet Algorithm[J]. Laser & Optoelectronics Progress, 2022, 59(16): 1615002
    Download Citation