3DRM: Pair-wise Relation Module for 3D Object Detection

Introduction

3DRM is a plug-and-play module which can be applied to different 3D detection frameworks to detect 3D objects more accurately and robustly. We implement two applications on proposal-based methods and vote-based methods: OSegNet (Proposal generation based on over-segmentation and backbones based on PointCNN) and VoteNet. OSegNet is trained and evaluated on S3DIS dataset, and VoteNet is trained and evaluated on ScanNetV2 and SUN RGB-D dataset. For more information, please refer to the paper.

Citation

If you find this project or this paper useful for your research, please cite:

@article{lan20213drm,
  title={3DRM: Pair-wise Relation Module for 3D Object Detection},
  author={Lan, Yuqing and Duan, Yao and Shi, Yifei and Huang, Hui and Xu, Kai},
  journal={Computers \& Graphics},
  year={2021},
  publisher={Elsevier}
}

Usage

3DRM can be widely applicable on different popular 3D detection frameworks. For further information about our implemention on two frameworks, please refer to the sub-directories: OSegNet-RM and VoteNet-RM.

Acknowledgemets

This code largely benefits from excellent works PointCNN and VoteNet and repositories, please also consider cite PointCNN and VoteNet and if you use this code.