EcoSplat: Efficiency-controllable Feed-forward 3D Gaussian Splatting from Multi-view Images

Jongmin Park^1* · Minh-Quan Viet Bui^1* · Juan Luis Gonzalez Bello¹ · Jaeho Moon¹ · Jihyong Oh^2† · Munchurl Kim^1†
1KAIST, South Korea, ²Chung-Ang University, South Korea
*Co-first authors (equal contribution), †Co-corresponding authors

CVPR 2026

Paper | Project Page | Code

📧 News

• March 09, 2026: Released inference code and pre-trained models.
• Jan 03, 2026: Initial repository created.

🚀 Code Release Plan

The full code and pretrained models will be released soon.

• ✅ Inference code
• ✅ Pretrained models
• ⬛ Training scripts
• ⬛ Dataset generation scripts

🛠️ Installation

Our code is developed using PyTorch 2.5.1, CUDA 12.4, and Python 3.11.

git clone https://github.com/KAIST-VICLab/EcoSplat.git
cd EcoSplat

conda create -y -n ecosplat python=3.11
conda activate ecosplat
bash setup.sh

📦 Model Zoo

Our pre-trained models are hosted on Hugging Face 🤗.

We assume the downloaded weights are located in the pretrained_weights directory.

📂 Datasets

Please refer to DATASETS.md for dataset preparation.

💻 Running the Code

Evaluation

To evaluate EcoSplat on RealEstate10K, run the following command. You can adjust the primitive_ratio as needed.

# RealEstate10K (enable test.align_pose=true if using evaluation-time pose alignment)
python -m src.main +experiment=ecosplat/re10k mode=test wandb.name=re10k \
    dataset/view_sampler@dataset.re10k.view_sampler=evaluation \
    dataset.re10k.view_sampler.index_path=assets/evaluation_index_re10k_small_16views.json \
    checkpointing.load=./pretrained_weights/ecosplat-stage2-re10k.ckpt \
    model.encoder.primitive_ratio=<PRIMITIVE_RATIO> \ 
    test.save_image=true test.align_pose=true \ 
    test.output_path=<YOUR_OUTPUT_PATH>

🙏 Acknowledgements

This project is built upon these excellent repositories: SPFSplat, NoPoSplat, pixelSplat, DUSt3R, and CroCo. We thank the original authors for their excellent work.

🌱 Citation

@article{park2025ecosplat,
  title={EcoSplat: Efficiency-controllable Feed-forward 3D Gaussian Splatting from Multi-view Images},
  author={Park, Jongmin and Bui, Minh-Quan Viet and Bello, Juan Luis Gonzalez and Moon, Jaeho and Oh, Jihyong and Kim, Munchurl},
  journal={arXiv preprint arXiv:2512.18692},
  year={2025}
}