Acta Optica Sinica, Volume. 42, Issue 15, 1515001(2022)

Triplet Network Based on Dynamic Feature Attention for Object Tracking

Zishuo Zhang1,2, Yong Song1,2、*, Xin Yang1,2, Yufei Zhao1,2, and Ya Zhou1,2
Author Affiliations
  • 1School of Optics and Photonics, Beijing Institute of Technology, Beijing 100081, China
  • 2Beijing Key Laboratory for Precision Optoelectronic Measurement Instrument and Technology, Beijing 100081, China
  • show less
    Figures & Tables(11)
    Schematic diagram of DFA-TriNet
    Structural diagram of DFA_update module
    Structural diagram of multilayer perceptual mixer block
    Structural diagram of DFA_cross module
    Experimental results based on OTB100 dataset. (a) Precision; (b) success rate
    Experimental results of different attribute sequences based on OTB100 dataset. (a1)(a2) Background clutters; (b1)(b2) deformation; (c1) (c2) fast motion; (d1) (d2) illumination variation; (e1)(e2) in-plane rotation; (f1)(f2) low resolution; (g1)(g2) motion blur; (h1) (h2) occlusion; (i1) (i2) out-of-plane rotation; (j1)(j2) out-of-view; (k1)(k2) scale variation
    Actual tracking results of proposed algorithm and comparison algorithms in video sequences with different attributes. (a) Basketball; (b) Bird1; (c) Board; (d) Ironman; (e) Soccer
    • Table 1. Positive and negative sample allocation of training data in video sequence with DFA-TriNet

      View table

      Table 1. Positive and negative sample allocation of training data in video sequence with DFA-TriNet

      Sample typeInitial frame template image(within 100 frames)Update template image(within 5 frames)
      Positive sampleSameSame

      Negative sample

      (equally distributed)

      SameDifferent
      DifferentSame
      DifferentDifferent
    • Table 2. Experimental results for updating parameter based on VOT2018 dataset

      View table

      Table 2. Experimental results for updating parameter based on VOT2018 dataset

      k12345678910
      EAO0.4520.4660.4630.4620.4690.4680.4650.4640.4650.462
    • Table 3. Results of ablation experiment based on VOT2018 dataset

      View table

      Table 3. Results of ablation experiment based on VOT2018 dataset

      IndexSiamRPN++SiamRPN+++ BASiamRPN+++ BA+UBSiamRPN+++ BA+UB+DFA
      EAO0.4140.4480.4590.469
      Tracking speed /(frame·s-135443837
    • Table 4. Experimental results based on VOT2018 dataset

      View table

      Table 4. Experimental results based on VOT2018 dataset

      IndexDaSiamRPNUPDTSiamRPNATOMUpdateNetSiamRPN++SiamBANSiamAttnProposed
      EAO0.3260.3780.3830.4010.4030.4140.4520.4700.469
      Accuracy0.5690.5360.5860.5900.5830.6000.5970.6300.614
      Robustness0.3370.1840.2760.2030.2250.2340.1780.1600.122
      Tracking speed /(frame·s-1590.438305535443337
    Tools

    Get Citation

    Copy Citation Text

    Zishuo Zhang, Yong Song, Xin Yang, Yufei Zhao, Ya Zhou. Triplet Network Based on Dynamic Feature Attention for Object Tracking[J]. Acta Optica Sinica, 2022, 42(15): 1515001

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Machine Vision

    Received: Jan. 13, 2022

    Accepted: Mar. 7, 2022

    Published Online: Aug. 4, 2022

    The Author Email: Yong Song (yongsong@bit.edu.cn)

    DOI:10.3788/AOS202242.1515001

    Topics