Chinese Journal of Liquid Crystals and Displays, Volume. 40, Issue 6, 867(2025)

3D hand reconstruction method based on adaptive occlusion recovery and topology-pose bidirectional perception

Jia LIU1, Nanxuan HUANG1, Dapeng CHEN1、*, and Lina WEI2
Author Affiliations
  • 1School of Automation, Nanjing University of Information Science & Technology, Nanjing 210044, China
  • 2School of Computer and Computing Science, Hangzhou City University, Hangzhou 310015, China
  • show less
    Figures & Tables(13)
    Overall framework of the proposed method
    Specific structure diagram of the adaptive occlusion recovery module
    Specific structure diagram of the topology-pose bidirectional perception module
    Visualization results of the ablation experiment of the adaptive occlusion recovery module
    Visualization results of the ablation experiment of the topology-pose bidirectional perception module
    Visualization results of the comparison experiments on the FreiHAND dataset
    Reconstruction visualization results of the Frei-HAND dataset
    Visualization results of the comparison experiments on the InterHand2.6M dataset
    Reconstruction visualization results of the Inter-Hand2.6M dataset
    • Table 1. Ablation experiment results of the adaptive occlusion recovery module

      View table
      View in Article

      Table 1. Ablation experiment results of the adaptive occlusion recovery module

      方法PA-MPJPEPA-MPVPE
      注意力机制7.27.8
      注意力机制+Mask6.36.7
      注意力机制+Mask+Lregion5.76.1
    • Table 2. Ablation experiment results of the topology-pose bidirectional perception module

      View table
      View in Article

      Table 2. Ablation experiment results of the topology-pose bidirectional perception module

      方法PA-MPJPEPA-MPVPE
      静态拓扑感知(Ⅰ)7.37.5
      (Ⅰ)+动态姿态感知(Ⅱ)6.77.2
      (Ⅱ)+双向感知交互(Ⅲ)6.16.5
      (Ⅲ)+全局特征交互(Ⅳ)5.76.1
    • Table 3. Comparison experiment results on the FreiHAND dataset

      View table
      View in Article

      Table 3. Comparison experiment results on the FreiHAND dataset

      方法MPJPE/mmMPVPE/mmPA-MPJPE/mmPA-MPVPE/mmF@5 mmF@15 mmFPS
      I2L-MeshNet1915.316.87.47.60.6810.97372
      METRO2214.315.56.76.80.7170.98127
      Mesh Graphormer1613.514.76.36.50.6870.98324
      Deformer2613.314.26.26.40.7430.984-
      PointHMR3513.114.96.16.60.7200.984-
      Ours12.513.65.76.10.7650.98639
    • Table 4. Comparison experiment results on the Inter-Hand2.6M dataset

      View table
      View in Article

      Table 4. Comparison experiment results on the Inter-Hand2.6M dataset

      方法MPJPEMPVPE
      IntagHand 98.799.03
      MeMaHand 228.658.89
      ACR 238.098.29
      Zuo308.348.51
      Ours8.078.22
    Tools

    Get Citation

    Copy Citation Text

    Jia LIU, Nanxuan HUANG, Dapeng CHEN, Lina WEI. 3D hand reconstruction method based on adaptive occlusion recovery and topology-pose bidirectional perception[J]. Chinese Journal of Liquid Crystals and Displays, 2025, 40(6): 867

    Download Citation

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

    Category:

    Received: Dec. 9, 2024

    Accepted: --

    Published Online: Jul. 14, 2025

    The Author Email: Dapeng CHEN (dpchen@nuist.edu.cn)

    DOI:10.37188/CJLCD.2024-0345

    Topics