Home

Awesome

:book: 3DGS-DET: Empower 3D Gaussian Splatting with Boundary Guidance and Box-Focused Sampling for 3D Object Detection

<p align="center"> <small> 🔥The first work to introduce 3D Gaussian Splatting into 3D Object Detection. ⭐Star 3DGS-DET. Thanks🔥 </small> </p>

[Paper]   <br>

<!-- > [Yang Cao](https://yangcaoai.github.io/), Yihan Zeng, [Hang Xu](https://xuhangcn.github.io/), [Dan Xu](https://www.danxurgb.net) <br> --> <!-- > The Hong Kong University of Science and Technology, Huawei Noah's Ark Lab -->

Yang Cao*, Yuanliang Ju*, Dan Xu <br> The Hong Kong University of Science and Technology<br>

:triangular_flag_on_post: Updates

☑ Being <a href="https://huggingface.co/papers?date=2024-10-03" style="color: #FF5733; font-weight: bold;">Top-5</a> in Hugging Face Daily Papers!

☐ The code and data will be released within a month of the paper's acceptance. Please stay tuned.

☑ Our paper 3DGS-DET is released, check out it on arXiv.

Table of Contents

Boundary Guidance

<img src="assets/fig1.png">

Pipeline of 3DGS for 3D Object Detection

<img src="assets/fig2.png">

Methods

<img src="assets/fig3.png">

Detection Samples

<img src="assets/fig4.png">

Guidance from Different Priors

<img src="assets/fig5.png">

Rendered Images

<img src="assets/fig6.png">

BibTeX

Cite 3DGS-DET by:


@misc{3dgsdet,
      title={3DGS-DET: Empower 3D Gaussian Splatting with Boundary Guidance and Box-Focused Sampling for 3D Object Detection}, 
      author={Yang Cao and Yuanliang Ju and Dan Xu},
      year={2024},
      eprint={2410.01647},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2410.01647}, 
}

Contact

If you have any question or collaboration needs, please email yangcao.cs@gmail.com.

Acknowledgement

3DGS-DET is implemented based on 3DGS and MMDetection3D. We appreciate their great codes.