Official repo for CoRL 2021 paper Correspondence-Free Point Cloud Registration with SO(3)-Equivariant Implicit Shape Representations (link)
After creating a virtual environment with python 3.8, inside the virtual environment, do:
python setup.py build_ext --inplace- Install torch-batch-svd
The preprocessed ModelNet40 dataset can be downloaded at this Google Drive link. It is processed by this repo to obtain water-tight meshes and occupancy value for points in the space, which are not available in the original ModelNet40 dataset (mentioned in the OccNet repo). Extract the files and create a symbolic link named ModelNet40_install under the root of this repo.
Examples are given in the files run_train.sh and run_test.sh.
If this work is helpful for your research, please consider citing the original authors' work:
@inproceedings{zhu2022correspondence,
title={Correspondence-free point cloud registration with SO (3)-equivariant implicit shape representations},
author={Zhu, Minghan and Ghaffari, Maani and Peng, Huei},
booktitle={Conference on Robot Learning},
pages={1412--1422},
year={2022},
organization={PMLR}
}