Geometry Sharing Network for 3D Point Cloud Classification and Segmentation. AAAI 2020

Mingye Xu, Zhipeng Zhou, Yu Qiao.

Overview

Futher information please contact Mingye Xu (my.xu@siat.ac.cn)

Citation

Please cite this paper if you want to use it in your work,

@misc{xu2019geometry,
    title={Geometry Sharing Network for 3D Point Cloud Classification and Segmentation},
    author={Mingye Xu and Zhipeng Zhou and Yu Qiao},
    year={2019},
    eprint={1912.10644},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}

Installation

Requirements

• Linux (tested on Ubuntu 14.04/16.04)
• Python 3.5+
• PyTorch 1.0

Install

Install this library by running the following command:

cd OP
python setup.py install
cd ../

Usage

Point Cloud Classification

• Run the training script:

python main.py 

• Run the evaluation script :

python main.py --eval True --model_path 'pretrained/model_1024_92.9.t7'

Other information

We will release part segmentation code later. Due to the differences of models, please contact us by email if you need the classification model(2048 input points) with 93.3% accuracy.

Acknowledgement

This code is based on DGCNN and Pointnet2.Pytorch.