High-Quality Multi-View Partial Point Cloud for Registration

Liang Pan1      Zhongang Cai2 3      Haiyu Zhao2 3      Shuai Yi2 3      Ziwei Liu1     

Partial-to-Partial Point Cloud Registration is widely used in many real applications, such as 6D pose estimation and 3D reconstruction. It aims to register a source point cloud to a target point cloud by estimating a rigid transformation based on the overlapped areas between the two point clouds.

Details


Partial-to-Partial Point Cloud Registration evaluates the performance on partial-to-partial point cloud registration. This is a large subset of MVP, containing massive high-quality incomplete point cloud pairs, and each partial point cloud pair is generated from the same 3D CAD model. It contains 9,600 number of partial point clouds:

  • The Training set has 6,400 partial point cloud pairs, groundtruth correspondences, category labels and complete point clouds.

  • The Testing set has 1,200 transformed partial point cloud pairs, canonical partial point cloud pairs, groundtruth transformations, groundtruth correspondences, category labels and complete point clouds.

  • The ExtraTesting set has 2,000 transformed partial point cloud pairs, and category labels.

    We keep the groundtruth transformations, and you can evaluate your method by submitting your results on the Codalab website. The general pipeline is training your method on the Training set while evaluating on the Testing set. Finally, you can generate your results for the ExtraTest set by your best model.

Download

Citation

@inproceedings{pan2021variational,
  title={Variational Relational Point Completion Network},
  author={Pan, Liang and Chen, Xinyi and Cai, Zhongang and Zhang, Junzhe and Zhao, Haiyu and Yi, Shuai and Liu, Ziwei},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={8524--8533},
  year={2021}
}
          

Contact

If you have any question, please contact Liang Pan at liang.pan@ntu.edu.sg.