Optics and Precision Engineering, Volume. 32, Issue 6, 901(2024)

Object 6-DoF pose estimation using auxiliary learning

Minjia CHEN1...2, Shaoyan GAI1,2,*, Feipeng DA1,2, and Jian YU1,23,* |Show fewer author(s)
Author Affiliations
  • 1School of Automation, Southeast University, Nanjing20096, China
  • 2Key Laboratory of Measurement and Control of Complex Engineering Systems, Ministry of Education, Southeast University, Nanjing10096, China
  • 3Key Laboratory of Space Photoelectric Detection and Perception, Ministry of Industry and Information Technology, Nanjing University of Aeronautics and Astronautics, Nanjing211106, China
  • show less
    Figures & Tables(16)
    Network architecture of object 6-DoF pose estimation based on auxiliary learning
    Dual-branch for point cloud registration
    Local feature extractor
    Spatial information encoder
    Pipeline of fine pose estimation
    Illustration of point pair feature
    Qualitative results of pose estimation on YCB-Video Dataset
    Qualitative results of pose estimation on LineMOD Dataset
    • Table 1. Quantitative evaluation results on YCB-Video Dataset

      View table
      View in Article

      Table 1. Quantitative evaluation results on YCB-Video Dataset

      DenseFusion12REDE15PR-GCN23BiCo-Net16DCL-Net24Ours
      AUC<2 cmAUC<2 cmAUC<2 cmAUC<2 cmAUC<2 cmAUC<2 cm
      002 master chef can96.4100.095.1100.097.1100.096.2100.096.1100.096.3100.0
      003 cracker box95.599.596.399.797.6100.096.6100.096.499.496.3100.0
      004 sugar box97.5100.097.4100.098.3100.097.8100.098.1100.097.8100.0
      005 tomato soup can94.696.996.9100.095.397.695.798.195.897.795.798.2
      006 mustard bottle97.2100.096.7100.097.9100.098.0100.098.7100.097.8100.0
      007 tuna fish can96.6100.096.6100.097.6100.096.5100.097.4100.096.6100.0
      008 pudding box96.5100.096.4100.098.4100.097.5100.098.2100.097.7100.0
      009 gelatin box98.1100.097.8100.096.294.498.8100.098.9100.098.8100.0
      010 potted meat can91.393.192.094.296.699.193.094.593.194.793.094.7
      011 banana96.6100.097.099.798.5100.097.1100.098.1100.097.5100.0
      019 pitcher base97.1100.097.5100.098.1100.097.6100.098.099.897.5100.0
      021 bleach cleanser95.8100.094.2100.097.9100.096.6100.097.0100.096.2100.0
      024 bowl88.298.896.799.390.396.696.7100.097.3100.096.5100.0
      025 mug97.1100.097.0100.098.1100.097.0100.097.8100.096.7100.0
      035 power drill96.098.797.099.698.1100.097.099.998.0100.096.9100.0
      036 wood block89.794.691.098.396.0100.092.190.193.997.594.195.0
      037 scissors95.2100.094.5100.096.7100.092.299.587.698.391.199.8
      040 large marker97.5100.097.8100.097.9100.097.4100.097.899.897.4100.0
      051 large clamp72.979.277.380.787.593.394.798.395.798.694.998.6
      052 extra large clamp69.876.385.982.079.784.688.290.288.887.288.390.8
      061 foam brick92.5100.094.6100.097.8100.097.2100.097.5100.096.1100.0
      MEAN93.196.894.597.895.898.596.098.896.699.095.999.0
    • Table 2. Quantitative evaluation results on LineMOD Dataset

      View table
      View in Article

      Table 2. Quantitative evaluation results on LineMOD Dataset

      DenseFusion12REDE15PR-GCN23BiCo-Net16DCL-Net24Ours
      ape92.395.697.697.397.497.6
      ben.93.299.499.298.899.499.1
      cam94.499.699.499.699.899.8
      can93.199.598.499.399.999.7
      cat96.599.598.7100.0100.099.9
      drill.87.099.398.898.999.999.2
      duck92.397.098.998.798.498.8
      egg.99.8100.099.999.8100.099.8
      glue100.099.9100.099.899.999.0
      hole.86.998.699.499.2100.099.0
      iron97.099.398.5100.0100.099.7
      lamp95.399.399.299.799.599.7
      phone92.899.398.499.299.799.1
      MEAN94.398.998.999.399.599.4
    • Table 3. Quantitative evaluation results on LM-O Dataset

      View table
      View in Article

      Table 3. Quantitative evaluation results on LM-O Dataset

      PVN3D13PR-GCN23BiCo-Net16DCL-Net24Ours
      MEAN63.265.069.570.671.3
      ape33.940.255.656.752.6
      can88.676.283.280.287.1
      cat39.157.047.348.145.8
      drill.78.482.369.981.471.4
      duck41.930.058.344.656.3
      egg.80.968.278.183.686.1
      glue68.167.076.979.183.0
      hole.74.797.287.291.388.5
    • Table 4. Comparison on BOP Challenge 2022

      View table
      View in Article

      Table 4. Comparison on BOP Challenge 2022

      InputMethodYCB-VideoLM-O
      AUC<2 cmADD(-S)
      RGBCosyPose2889.8--
      GDR-Net2591.660.162.2
      PFA26-62.864.1
      ZebraPose2790.180.576.9
      RGB-DRCVPose3D2996.6-74.5
      Ours95.999.071.3
    • Table 5. Comparison of inference time on LineMOD dataset

      View table
      View in Article

      Table 5. Comparison of inference time on LineMOD dataset

      SegPERefineALL
      DenseFusion1230201060
      PR-GCN233083068
      BiCo-Net163045-75
      Ours3039-69
    • Table 6. Comparison of model parameters

      View table
      View in Article

      Table 6. Comparison of model parameters

      Main modelRefinerALL
      DCL-Net2450.310.560.8
      BiCo-Net16191.1-191.1
      Ours41.9-41.9
    • Table 7. Experiment results of different branches

      View table
      View in Article

      Table 7. Experiment results of different branches

      Pose RegC2MM2CAL-Netresults
      ××66.3
      ××65.9
      ×68.8
      ×71.3
    • Table 8. Experiment results of auxiliary learning network

      View table
      View in Article

      Table 8. Experiment results of auxiliary learning network

      LFESIECoarseC2Fresults
      ×-70.5
      ×-70.4
      -69.8
      -71.3
    Tools

    Get Citation

    Copy Citation Text

    Minjia CHEN, Shaoyan GAI, Feipeng DA, Jian YU. Object 6-DoF pose estimation using auxiliary learning[J]. Optics and Precision Engineering, 2024, 32(6): 901

    Download Citation

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

    Category:

    Received: Jul. 26, 2023

    Accepted: --

    Published Online: Apr. 19, 2024

    The Author Email: GAI Shaoyan (qxxymm@163.com), YU Jian (yujian@seu.edu.cn)

    DOI:10.37188/OPE.20243206.0901

    Topics