/SpinNet

Generalized 3D Surface Descriptor

MIT LicenseMIT

PWC License CC BY-NC-SA 4.0

SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration

This is the repository of SpinNet ([Arxiv report]), a conceptually simple neural architecture to extract local features which are rotationally invariant whilst sufficiently informative to enable accurate registration. For technical details, please refer to:

SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration
Sheng Ao*, Qingyong Hu*, Bo Yang, Andrew Markham, Yulan Guo.
[Paper] [Video] [Project page]

(1) Overview

(2) Results on Public Datasets

  • Comparisons with the State-of-the-arts.

  • Performance under Different Number of Sampled Points

  • Performance under Different Error Thresholds

(3) Generalization Performance

  • Generalization From 3DMatch to ETH

  • Generalization From KITTI to 3DMatch

  • Generalization From 3DMatch to KITTI

(4) Qualitative Results

Citation

If you find our work useful in your research, please consider citing:

@article{ao2020SpinNet,
  title={SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration},
  author={Ao, Sheng and Hu, Qingyong and Yang, Bo and Markham, Andrew and Guo, Yulan},
  journal={arXiv preprint arXiv:2011.12149},
  year={2020}
}

Updates

  • 25/11/2020: Initial release