Adaptive Graph Convolution for Point Cloud Analysis

This repository contains the implementation of AdaptConv for point cloud analysis.

Adaptive Graph Convolution (AdaptConv) is a point cloud convolution operator presented in our ICCV2021 paper. If you find our work useful in your research, please cite our paper.

preprint:

@article{zhou2021adaptive,
  title={Adaptive Graph Convolution for Point Cloud Analysis},
  author={Zhou, Haoran and Feng, Yidan and Fang, Mingsheng and Wei, Mingqiang and Qin, Jing and Lu, Tong},
  journal={arXiv preprint arXiv:2108.08035},
  year={2021}
}

Installation

• The code has been tested on one configuration:

  • PyTorch 1.1.0, CUDA 10.1

• Install required packages:

  • numpy
  • h5py
  • scikit-learn
  • matplotlib

Classification

classification.md

Part Segmentation

part_segmentation.md

Indoor Segmentation

sem_segmentation.md

Updates

• 09/30/2021: Updated code for part segmentation.
• 09/30/2021: Added code for S3DIS indoor segmentation.