Home

Awesome

PA-AUG + MD3D

This repository contains combinations of Part-Aware Data Augmentation for 3D Object Detection in Point Cloud (IROS 2021) and MD3D: Mixture-Density-based 3D Object Detection in Point Clouds (IEEE Access).

<p align="center"> <img src="docs/pa-aug.png" width="90%"> </p> <p align="center"> <img src="docs/md3d_architecture.png" width="90%"> </p>

Installation & Training

README_OpenPCDet.md

Results

<p align="center"> <img src="docs/result1.png" width="90%"> </p> <p align="center"> <img src="docs/result2.png" width="90%"> </p>

Citation

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

@inproceedings{choi2021part,
  title={Part-aware data augmentation for 3d object detection in point cloud},
  author={Choi, Jaeseok and Song, Yeji and Kwak, Nojun},
  booktitle={2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={3391--3397},
  year={2021},
  organization={IEEE}
}

@article{choi2022md3d,
  author={Choi, Jaeseok and Song, Yeji and Kim, Yerim and Yoo, Jaeyoung and Kwak, Nojun},
  journal={IEEE Access}, 
  title={MD3D: Mixture-Density-Based 3D Object Detection in Point Clouds}, 
  year={2022},
  volume={10},
  number={},
  pages={104011-104022},
  doi={10.1109/ACCESS.2022.3210108}
}

Acknowledgement