HyDRa

Unleashing HyDRa: Hybrid Fusion, Depth Consistency and Radar for Unified 3D Perception

Philipp Wolters, Johannes Gilg, Torben Teepe, Fabian Herzog, Anouar Laouichi, Martin Hofmann, Gerhard Rigoll

Introduction

This repository is an official implementation of HyDRa, our novel camera-radar fusion architecture for 3D perception.

News

• [2024/3/19] Official PyTorch implementation is coming soon.
• [2024/3/12] Our preprint is available on arXiv.

Getting Started

Model Zoo

3D Object Detection on nuScenes

Method	Backbone	NDS	mAP	Config	Checkpoint
HyDRa	R50	58.5	49.4	config	model
HyDRa	R101	61.7	53.6	config	model
HyDRa	V2-99	64.2	57.4	config	model

3D Semantic Occupancy Prediction on Occ3D

Method	Backbone	mIoU	Config	Checkpoint
FB-OCC	R50	39.1	config	model
HyDRa	R50	44.4	config	model

Citation

If this work is helpful for your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.

@article{wolters2024unleashing,
  title={Unleashing HyDRa: Hybrid Fusion, Depth Consistency and Radar for Unified 3D Perception},
  author={Wolters, Philipp and Gilg, Johannes and Teepe, Torben and Herzog, Fabian and Laouichi, Anouar and Hofmann, Martin and Rigoll, Gerhard},
  journal={arXiv preprint arXiv:2403.07746},
  year={2024}
}

Acknowledgements

We would like to thank other great open source projects:

• BEVDet, SoloFusion, FB-BEV , CRN, Occ3D