Hao Li, Sicheng Li, Xiang Gao, Abudouaihati Batuer, Lu Yu, Yiyi Liao
Overview: we introduce GIFStream, a novel 4D Gaussian representation enabling high quality representation and efficient compression
Immersive video offers a 6-Dof-free viewing experience, potentially playing a key role in future video technology. Recently, 4D Gaussian Splatting has gained attention as an effective approach for immersive video due to its high rendering efficiency and quality, though maintaining quality with manageable storage remains challenging. To address this, we introduce GIFStream, a novel 4D Gaussian representation using a canonical space and a deformation field enhanced with time-dependent feature streams. These feature streams enable complex motion modeling and allow efficient compression by leveraging their motion-awareness and temporal correspondence. Additionally, we incorporate both temporal and spatial compression networks for endto-end compression. Experimental results show that GIFStream delivers high-quality immersive video at 30 Mbps, with real-time rendering and fast decoding on an RTX 4090.
# Clone the repo.
git clone https://github.com/Hliii/GIFStream.git --recursive
cd GIFStream
# Make a conda environment
conda create --name GIFStream python=3.10
conda activate GIFStreamPlease install Pytorch first.
Then, you can install the extended gsplat library with GIFStream training, rendering and compression features.
pip install .If you want to do further development based on this framework, you can use following command to install Python packages in editable mode.
pip install -e . # (develop)Same as gsplat, we need to install some extra dependencies.
cd examples
pip install -r requirements.txt
cd ../third_party/MLEntropy
mkdir build
cd build
cmake ../cpp -DCMAKE_BUILD_TYPE=Release
make -jFor Neur3D dataset, please first download the dataset and the organization of files should be like this:
└── Neur3D/
├── coffee_martini/
│ ├── cam00.mp4
│ └── ...
├── cook_spinach/
│ ├── cam00.mp4
│ └── ...
└── ...Then preprocess the data using the script as below.
python dataset_process/n3d_video_process.py --root_dir your_path_to_neur3d_datasetWe provide a script that enables end-to-end compression-aware training and compression for videos containing several gops.
bash examples/benchmarks/multigop_gifstream.sh- Release code using gsplat and gscodec studio framework.
This project is bulit on gsplat, GScodec Studio, Scaffold-GS and DCVC-HEM. We thank all contributors for such great open-source projects.
Please cite our paper if you find this repository useful:
@misc{li2025gifstream4dgaussianbasedimmersive,
title={GIFStream: 4D Gaussian-based Immersive Video with Feature Stream},
author={Hao Li and Sicheng Li and Xiang Gao and Abudouaihati Batuer and Lu Yu and Yiyi Liao},
year={2025},
eprint={2505.07539},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2505.07539},
}