RGM: Robust Point Cloud Registration Framework Based on Deep Graph Matching (CVPR2021)

This repository provides code and data required to train and evaluate RGM. It represents the official implementation of the paper:

Robust Point Cloud Registration Framework Based on Deep Graph Matching

Kexue Fu, Shaolei Liu, Xiaoyuan Luo, Manning Wang

Instructions

This code has been tested on

• Python 3.6.10, PyTorch 1.2.0, CUDA 10.2, GeForce RTX 2080Ti/GeForce GTX 1080.

Requirements

To create a virtual environment and install the required dependences please run:

git clone https://github.com/fukexue/RGM.git
conda create -n RGM
conda activate RGM
pip install -r requirements.txt

in your working folder.

Note: If you want to get the same results as in the paper, install numpy.version=='1.19.2' and scipy.version=='1.5.0'.

Datasets and pretrained models

For ModelNet40, the data will be downloaded automatically.

For ShapeNet dataset, please download it from this link ShapeNet. Unzip, named 'shapenet_raw' and place it in the data folder.

We provide

• pretrained models on ModelNet40 in clear, nosie, and partial. you can download it from this link weight. Unzip and place it in the output folder.

Train

sh 1_experiment_train.sh

Eval

sh 1_experiment_eval.sh

Citation

If you find this code useful for your work or use it in your project, please consider citing:

@article{Fu2021RGM,
  title={Robust Point Cloud Registration Framework Based on Deep Graph Matching},
  author={Kexue Fu, Shaolei Liu, Xiaoyuan Luo, Manning Wang},
  journal={Internaltional Conference on Computer Vision and Pattern Recogintion (CVPR)},
  year={2021}
}

Acknowledgments

In this project we use (parts of) the official implementations of the followin works:

• DGCNN (feature)
• RPMNet (ModelNet preprocessing and evaluation)
• PCA-GM

We thank the respective authors for open sourcing their methods.