• Laser & Optoelectronics Progress
  • Vol. 57, Issue 2, 21510 (2020)
Dong Jifu1, Liu Chang1、*, Cao Fangwei1, Ling Yuan2, and Gao Xiang1
Author Affiliations
  • 1College of Information Sciences and Technology, Dalian Maritime University, Dalian, Liaoning 116026, China
  • 2Hualu Zhida Technology Co., Ltd., Dalian, Liaoning 116023, China
  • show less
    DOI: 10.3788/LOP57.021510 Cite this Article Set citation alerts
    Dong Jifu, Liu Chang, Cao Fangwei, Ling Yuan, Gao Xiang. Online Adaptive Siamese Network Tracking Algorithm Based on Attention Mechanism[J]. Laser & Optoelectronics Progress, 2020, 57(2): 21510 Copy Citation Text show less
    Framework of online adaptive siamese network based on attention mechanism
    Fig. 1. Framework of online adaptive siamese network based on attention mechanism
    Cropped (t-1)th frame
    Fig. 2. Cropped (t-1)th frame
    Spatial attention module
    Fig. 3. Spatial attention module
    Channel attention module
    Fig. 4. Channel attention module
    Plots of accuracy and average success rate on OTB50 dataset. (a) Plot of accuracy; (b) plot of average success rate
    Fig. 5. Plots of accuracy and average success rate on OTB50 dataset. (a) Plot of accuracy; (b) plot of average success rate
    Plots of accuracy and average success rate on OTB100 dataset. (a) Plot of accuracy; (b) plot of average success rate
    Fig. 6. Plots of accuracy and average success rate on OTB100 dataset. (a) Plot of accuracy; (b) plot of average success rate
    Average success rates of similar background attributes on OTB100 dataset
    Fig. 7. Average success rates of similar background attributes on OTB100 dataset
    Average success rates of deformation properties on OTB100 dataset
    Fig. 8. Average success rates of deformation properties on OTB100 dataset
    Expected average overlap rates on VOT2016 dataset
    Fig. 9. Expected average overlap rates on VOT2016 dataset
    Qualitative evaluation of different algorithms on seven challenging video sequences
    Fig. 10. Qualitative evaluation of different algorithms on seven challenging video sequences
    TypeSize /StrideNumber of filtersSize of exemplarSearch size
    Conv111×11/29659×59123×123
    Pool13×3/229×2961×61
    Conv25×525625×2557×57
    Pool23×3/212×1228×28
    Conv33×338410×1026×26
    Conv43×33848×824×24
    Conv53×32566×622×22
    Table 1. Parameters of SiamFC network
    TrackerEAOAccuracyRobustness
    AAM-Siam0.320.560.32
    Staple0.300.540.38
    DeepSRDCF0.280.520.33
    SiamFC0.280.550.38
    KCF0.190.490.57
    Table 2. Evaluation of accuracy and robustness on VOT2016 dataset
    Dong Jifu, Liu Chang, Cao Fangwei, Ling Yuan, Gao Xiang. Online Adaptive Siamese Network Tracking Algorithm Based on Attention Mechanism[J]. Laser & Optoelectronics Progress, 2020, 57(2): 21510
    Download Citation