SmartSplat: Feature-Smart Gaussians for Scalable Compression of Ultra-High-Resolution Images

🌐Anonymous Website

Raw Image info: 16320×10848, 189 MB

Table of Contents

1. Installation
2. Datasets
3. Benchmarking
4. Acknowledgement

Installation

conda create -n smartsplat python==3.12
conda activate smartsplat

# install torch
pip install torch==2.5.1 torchvision==0.20.1 --index-url https://download.pytorch.org/whl/cu124

pip install setuptools==78.0.1

pip install -r requirements.txt


# install Gaussian Rasterization
cd submodules/fused-ssim
pip install -e .
cd ../gsplat
pip install -e .
cd ../gsplat2d
pip install -e .
cd ../simple-knn-2d-qr
pip install -e .

Datasets

You can download the DIV8K dataset from huggingface, and the DIV16K dataset will be made publicly available after the paper is accepted.

Benchmarking

This codebase integrates multiple GS-based image representation methods, including GaussianImage, ImageGS, 3DGS, and LIG.

All our experiments were conducted on the A800 cluster. You can find the relevant run scripts in the slurm folder, and the experimental test logs are available in the slurm_logs folder.

Acknowledgement

We thank the authors of the following repositories for their open-source code:

• GaussianImage
• Gsplat