Mingyang Zhao · Gaofeng Meng · Dong-Ming Yan
Project Page | Paper | Poster
This repository contains the official implementation of our ICLR 2025 paper "Occlusion-aware Non-Rigid Point Cloud Registration via Unsupervised Neural Deformation Correntropy".
Towards Non-rigid or Deformable Registration of Point Clouds and Surfaces
Please give a star if you find this repo useful 🤡
The code is based on PyTorch implementation, and tested on the following environment dependencies:
- Linux (tested on Ubuntu 22.04.1)
- Python 3.9.19
- torch=='1.12.1+cu113'
We recommend using Miniconda to set up the environment.
- conda create -n oar python=3.9
- conda activate oar
- pip install torch==1.10.1+cu111 torchvision==0.11.2+cu111 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
- conda install -c fvcore -c iopath -c conda-forge fvcore iopath
- conda install pytorch3d
If you want the torch version match the pytorch3d version, please use conda list to check the corresponding Version, and then re-setup the torch, such as
- pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1 --extra-index-url https://download.pytorch.org/whl/cu113
Finally, setup other libraries:
- pip install -r requirements.txt
- cd src
- python test_OAR.py
The deformed point clouds are save in the subdirectory save_deformed of the directory data .
If you have any problem, please contact us via migyangz@gmail.com. We greatly appreciate everyone's feedback and insights. Please do not hesitate to get in touch!
Please give a citation of our work if you find it useful:
@inproceedings{zhao2025oareg,
title={Occlusion-aware Non-Rigid Point Cloud Registration via Unsupervised Neural Deformation Correntropy},
author={Mingyang Zhao, Gaofeng Meng, Dong-Ming Yan},
booktitle={International Conference on Learning Representations},
year={2025}
}Our work is inspired by several outstanding prior works, including DPF, NSFP, NDP, and others. We would like to acknowledge and express our deep appreciation to the authors of these remarkable contributions.
OAReg is under AGPL-3.0, so any downstream solution and products (including cloud services) that include OAReg code inside it should be open-sourced to comply with the AGPL conditions. For learning purposes only and not for commercial use. If you want to use it for commercial purposes, please contact us first.