Indirect Registration

This repository contains the implementation of our RAL 2022 paper: Indirect Point Cloud Registration: Aligning Distance Fields using a Pseudo Third Point Set.

Yijun Yuan, Andreas Nüchter

Preprint | Supplementary

Install

conda create -n ifr_exp python=3.6
conda activate ifr_exp

pip install -r requirements.txt

updates

• [30.6.2022] upload a simple setup.py.

  not necessary, but for easy embeding, just python setup.py install and will install a IFR package, se_math package contains IFR.ifr, IFR.utils and etc.

Quick Demo & How to use

For fast run, we provide a jupyter notebook for registration demo. Please find ./demo/test_toysample.ipynb

You may need jupyter nbextension enable --py --sys-prefix widgetsnbextension to enable the ipyvolume plot in notebook

Then jupyter notebook

So in that notebook, you can find the usage

from ifr import IFR
ifr_model = IFR(*param)
T = ifr.register(p0,p1)[0,:,:] # register p1 (source) to p0 (target)

Dataset & Test

The following is related to experiment in paper.

The test script are highly compatible to PointNetLK-Re tests. Please follow repo for preparing dataset.

Afterwards, for testing, please find ./scripts/.

source ./scripts/modelnet.sh
source ./scripts/shapenet.sh
source ./scripts/3dmatch.sh

Citation

If you find this work interesting, please cite us:

@article{yuan2022indirect,
  title={Indirect Point Cloud Registration: Aligning Distance Fields using a Pseudo Third Point Set },
  author={Yuan, Yijun and N{\"u}chter, Andreas},
  journal={IEEE Robotics and Automation Letters},
  year={2022},
  publisher={IEEE}
}

Acknowlegement

This code is mostly on top of PointNetLK-Re. We thank for the kind response of Xueqian Li.