Home

Awesome

<h2 align="center">Fully Explicit Dynamic Gaussian Splatting</h2> <p align="center"> <a href="https://leejunoh.com/"><strong>Junoh Lee</strong></a> · <strong>ChangYeon Won</strong> · <strong>Hyunjun Jung</strong> · <a href="https://ihbae.com/"><strong>Inhwan Bae</strong></a> · <a href="https://scholar.google.com/citations?user=Ei00xroAAAAJ"><strong>Hae-Gon Jeon</strong></a> <br> NeurIPS 2024 </p> <p align="center"> <a href="https://leejunoh.com/Ex4DGS/"><strong><code>Project Page</code></strong></a> <a href="https://neurips.cc/virtual/2024/poster/94164"><strong><code>NeurIPS Paper</code></strong></a> <a href="http://arxiv.org/abs/2410.15629"><strong><code>Arxiv Paper</code></strong></a> <a href="https://github.com/juno181/Ex4DGS"><strong><code>Source Code</code></strong></a> <a href="#-citation"><strong><code>Related Works</code></strong></a> </p> <div align='center'> <br><img src="img/Ex4DGS-thumbnail.gif" width=70%> <!--<img src="img/eigentrajectory-model.svg" width=70%>--> <br>A novel view synthesis result of Ex4DGS. </div>

<br>Summary: 4D Gaussian Splatting with static & dynamic separation using an incrementally extensible, keyframe-based model

<br>

Contents

  1. Setup
  2. Preprocess Datasets
  3. Training
  4. Evaluation
  5. Pretrained models

Setup

Environment Setup

Clone the source code of this repo.

git clone https://github.com/juno181/Ex4DGS.git
cd Ex4DGS

Installation through pip is recommended. First, set up your Python environment:

conda create -n Ex4DGS python=3.9
conda activate Ex4DGS

Make sure to install CUDA and PyTorch versions that match your CUDA environment. We've tested on RTX 4090 GPU with PyTorch version 2.12. Please refer https://pytorch.org/ for further information.

pip install torch

The remaining packages can be installed with:

pip install --upgrade setuptools cython wheel
pip install -r requirements.txt
<!-- Our default, provided install method is based on Conda package and environment management: ```shell conda env create --file environment.yml conda activate Ex4DGS ``` -->

Preprocess Datasets

For dataset preprocessing, we follow STG.

Neural 3D Video Dataset

First, download the dataset from here. You will need colmap environment for preprocess. To setup dataset preprocessing environment, run scrips:

./scripts/env_setup.sh

To preprocess dataset, run script:

./scripts/preprocess_all_n3v.sh <path to dataset>

Technicolor dataset

Download the dataset from here. To setup dataset preprocessing environment, run scrips:

./scripts/preprocess_all_techni.sh <path to dataset>

Please refer STG for further information.

Training

Run command:

python train.py --config configs/<some config name>.json --model_path <some output folder>  --source_path <path to dataset>

Evaluation

Run command:

python render.py --model_path <path to trained model>  --source_path <path to dataset> --skip_train --iteration <trained iter>

Pretrained models

We provide pretrained models in release.

📖 Citation

<!-- If you find this code useful for your research, please cite our trajectory prediction papers :) -->

🛍️ Ex4DGS (NeurIPS'24) 🛍️ |

@inproceedings{lee2024ex4dgs,
  title={Fully Explicit Dynamic Guassian Splatting},
  author={Lee, Junoh and Won, ChangYeon and Jung, Hyunjun and Bae, Inhwan and Jeon, Hae-Gon},
  booktitle={Proceedings of the Neural Information Processing Systems},
  year={2024}
}
<br>