Skip to content

Latest commit

 

History

History
89 lines (65 loc) · 2.92 KB

README.md

File metadata and controls

89 lines (65 loc) · 2.92 KB
Raw

Reliable Inlier Evaluation for Unsupervised Point Cloud Registration

by Yaqi Shen, Le Hui, Haobo Jiang, Jin Xie and Jian Yang, details are in paper.

Introduction

This repository contains the source code and pre-trained models for RIENet (published on AAAI 2022).

Usage

  1. Requirement:

    • Hardware: GeForce_RTX_2080_Ti
    • Software: PyTorch>=1.7.1, Python3, CUDA>=11.0, scipy>=1.5.4, tensorboardX, h5py, tqdm, easydict, yaml, sklearn, plyfile, MinkowskiEngine>=0.5

  2. Clone the repository and build the ops:

    git clone https://github.com/supersyq/RIENet.git
cd RIENet
cd pointnet2 && python setup.py install && cd ../

  3. Datasets

    (1) ModelNet40

    (2) 7Scenes

    7scene
├── 7-scenes-chess
│   ├── cloud_bin_0.info.txt
│   ├── cloud_bin_0.ply
|   ├── ...
├── 7-scenes-fire
├── ...

    (3) ICL-NUIM

    (4) KITTI

    sequences
├── 00
│   ├── velodyne
│   ├── calib.txt
├── 01
├── ...

  4. Train:

    • Modify the 'data_file_test', 'data_file', 'gaussian_noise', 'dataset-path', 'root', '' specified in folder 'config' and then do training:

      CUDA_VISIBLE_DEVICES=0 python main.py ./config/train.yaml
CUDA_VISIBLE_DEVICES=0 python main.py ./config/train7.yaml
CUDA_VISIBLE_DEVICES=0 python main.py ./config/train-icl.yaml
CUDA_VISIBLE_DEVICES=0 python main.py ./config/train-k.yaml

  5. Test:

    • We provide pretrained models in ./pretrained, please modify eval specified in folder 'config' and then do testing:

      CUDA_VISIBLE_DEVICES=0 python main.py ./config/train.yaml
CUDA_VISIBLE_DEVICES=0 python main.py ./config/train7.yaml
CUDA_VISIBLE_DEVICES=0 python main.py ./config/train-icl.yaml
CUDA_VISIBLE_DEVICES=0 python main.py ./config/train-k.yaml

Citation

If you find the code or trained models useful, please consider citing:

@inproceedings{shen2022reliable,
  title={Reliable Inlier Evaluation for Unsupervised Point Cloud Registration},
  author={Shen, Yaqi and Hui, Le and Jiang, Haobo and Xie, Jin and Yang, Jian},
  booktitle={AAAI},
  year={2022}
}

Acknowledgement

Our code refers to DCP, RPMNet, FMR, DeepGMR, and HRegNet. We want to thank the above open-source projects.