H3D-DGS

KO | EN

gitlite — search

#javascript #typescript #llm #golang #machine-learning #linux #java #windows #awesome #cli #awesome-list #pytorch #security #openai

(★ 19)

[2025‘NeurIPS] Official Code for H3D-DGS

H3D-DGS 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# H3D-DGS: Exploring Heterogeneous 3D Motion Representation for Deformable 3D Gaussian Splatting

<div align="center">
<a href="https://openreview.net/forum?id=6VoDizmIoY"><img src='https://img.shields.io/badge/arXiv-2408.13036-b31b1b.svg'></a> &nbsp;
<a href="https://hebing-sjtu.github.io/H3D-DGS-page/"><img src='https://img.shields.io/badge/Project-In%20Progress-brightgreen.svg'></a> &nbsp;

_**[Bing He<sup>*;</sup>](https://github.com/hebing-sjtu), 
Yunuo Chen<sup>*</sup>, 
Guo Lu, 
Xie Rong,
Li Song <sup>&dagger;</sup>, 
Wenjun Zhang**_
<br><br>
Shanghai Jiao Tong University

<h2 align="center">🌟 <b>NeurIPS 2025</b> 🌟</h2>

</div>

![image](https://raw.githubusercontent.com/hebing-sjtu/H3D-DGS/master/assets/teaser.png)
## 🔆 Introduction

### H3D Control Points

Brand-new motion discrete representation tailored for neural volume rendering. Marry neural representation with traditional graphics. H3D Control Points inherit 3D motion from 2D optical prior, thus effective constraint and fast convergence. No fear for complex motion.

![image](https://raw.githubusercontent.com/hebing-sjtu/H3D-DGS/master/assets/H3D.png)

### Streaming Pipeline

A straightforward but effective workflow enable **streaming** dynamic 3D reconstruction utilizing H3D points and 3D Gaussians.

![image](https://raw.githubusercontent.com/hebing-sjtu/H3D-DGS/master/assets/workflow.png)

## 🧰 Installation

assume that you have already installed nvcc version>=12.1,
or else you should follow the offical document to install CUDA driver as well as Nvidia toolkit.
https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html

First clone this repository.
```
# Clone this repo(pytorch)
git clone https://github.com/hebing-sjtu/H3D-DGS.git --recursive
# or
git clone https://github.com/hebing-sjtu/H3D-DGS.git
git submodule update --init --recursive
```

Then install the dependence of this repo
```
# Install this repo
conda create -n h3d-dgs python=3.10 -y
conda activate h3d-dgs
pip install torch==2.5.0 torchvision==0.20.0 torchaudio==2.5.0 --index-url https://download.pytorch.org/whl/cu121
pip install -r requirements.txt
```

Compile the rasterization as well as knn
```
# Install rasterization
cd submodules/rasterization
python setup.py install
pip install .
cd ../..
# Install KNN with CUDA support
cd submodules/KNN
make && make install
cd ../..
```

Optionally compile the kmeans tool for control points prune
```
# Install KMeans with CUDA support(optional)
cd submodules/KMeans
python setup.py install
pip install .
cd ../..
```

```

## 📜 Citation

If you find our work useful, please consider citing:

```
@inproceedings{heh3d,
  title={H3D-DGS: Exploring Heterogeneous 3D Motion Representation for Deformable 3D Gaussian Splatting},
  author={He, Bing and Chen, Yunuo and Lu, Guo and Wang, Qi and Gu, Qunshan and Xie, Rong and Song, Li and Zhang, Wenjun},
  booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems}
}
```

H3D-DGS

H3D-DGS 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

H3D-DGS

H3D-DGS 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README