Dynamic Graph CNN for Learning on Point Clouds

We propose a new neural network module dubbed EdgeConv suitable for CNN-based high-level tasks on point clouds including classification and segmentation. EdgeConv is differentiable and can be plugged into existing architectures.

[Project] [Paper]

Overview

DGCNN is the author's re-implementation of Dynamic Graph CNN, which achieves state-of-the-art performance on point-cloud-related high-level tasks including category classification, semantic segmentation and part segmentation.

Further information please contact Yue Wang and Yongbin Sun.

Requirements

• TensorFlow

Point Cloud Classification

• Run the training script:

python train.py

• Run the evaluation script after training finished:

python evalutate.py

Citation

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

@article{dgcnn,
  title={Dynamic Graph CNN for Learning on Point Clouds},
  author={Yue Wang, Yongbin Sun, Ziwei Liu, Sanjay E. Sarma, Michael M. Bronstein, Justin M. Solomon},
  journal={arXiv preprint arXiv:1801.07829},
  year={2018}
}

License

MIT License

Acknowledgement

This code is heavily borrowed from PointNet.