PRIN/SPRIN: On Extracting Point-wise Rotation Invariant Features (TPAMI 2021)

News

• [2022-03] Our new sim-to-real 9D pose estimation method CPPF that leverages SPRIN SE(3)-invariant features is accepted to CVPR 2022.

Overview

This repository is the Pytorch implementation of PRIN/SPRIN: On Extracting Point-wise Rotation Invariant Features (TPAMI 2021), which is an improved version of our previous work.

ShapeNet Part Segmentation results

Feature Matching Results for PRIN

Dataset and pretrained weights

• Download ShapeNet part segmentation dataset from https://shapenet.cs.stanford.edu/media/​shapenet_part_seg_hdf5_data.zip
• Pretrained weights for SPRIN can be downloaded from https://drive.google.com/file/d/1ZwN0bJ3UgCJheqYcUrEZyHIH8tEgk3cR/view?usp=sharing
• Pretrained weights for PRIN can be downloaded from https://drive.google.com/file/d/116c80nvL6jAK4T75d5IMudgI8KRcMFa0/view?usp=sharing

Train and Test for PRIN/SPRIN

Please refer to README under folder prin and sprin.

License

MIT

References

Our paper is available on https://arxiv.org/pdf/2102.12093.pdf.

Citation

@article{you2021prinsprin,
  title={PRIN/SPRIN: On Extracting Point-wise Rotation Invariant Features},
  author={Yang You and Yujing Lou and Ruoxi Shi and Qi Liu and Yu-Wing Tai and Lizhuang Ma and Weiming Wang and Cewu Lu},
  journal={arXiv preprint arXiv:2102.12093},
  year={2021}
}