Point Cloud Instance Segmentation with Semi-supervised Bounding-Box Mining

Introduction

This repository is inference code release for our T-PAMI 2021 paper (arXiv report here).

Citation

If you find our work useful in your research, please consider citing:

@inproceedings{YongbinLiao2022PointCI,
  title={Point Cloud Instance Segmentation with Semi-supervised Bounding-Box Mining},
  author={Yongbin Liao and Hongyuan Zhu and Yanggang Zhang and Chuangguan Ye and Tao Chen and Jiayuan Fan},
  year={2022}
}

Installation

Install Pytorch and Tensorflow (for TensorBoard). It is required that you have access to GPUs.The code is tested with Ubuntu 18.04, Pytorch v1.8, TensorFlow v1.15.

Compile the CUDA layers for PointNet++, which we used in the backbone network:

cd pointnet2
python setup.py install

Install the following Python dependencies (with pip install):

matplotlib
opencv-python
plyfile
'trimesh>=2.35.39,<2.35.40'
'networkx>=2.2,<2.3'
scipy

Install the following Python dependencies (with conda install):

conda install -c conda-forge point_cloud_utils

Dataset preparation

We follow the VoteNet codebase for preprocessing our data. The instructions for preprocessing SUN RGB-D are here and ScanNet are here.

Run eval

You can download pre-trained models HERE. Unzip the file under the project root path (/path/to/project/checkpoint_dir) and then run:

python eval.py