[NeurIPS 2025] Projection-Manifold Regularized Latent Diffusion for Robust General Image Fusion

This repository is the official implementation of the NeurIPS 2025 paper: "Projection-Manifold Regularized Latent Diffusion for Robust General Image Fusion".

Paper | Poster

📢 News

• Code and demo data are released.

🛠️ Installation

Clone this repository:

git clone https://github.com/Leiii-Cao/PDFuse.git
cd PDFuse

Create and activate the conda environment:

conda create -n PDFuse python=3.11.5
conda activate PDFuse

Install the required packages:

pip install -r requirements.txt
pip install --force-reinstall "huggingface-hub==0.23.3"

🚀 Inference

Please download the Stable Diffusion v1.5 model from Hugging Face:

• Stable-Diffusion-v1-5 The default model path used by the code is:

./pretrained/stable-diffusion-v1-5/

The repository includes several demo image pairs under data/.

bash test.sh

bash test_E.sh

📝 Citation

If our work assists your research, feel free to give us a star ⭐ or cite us using:

@article{cao2026projection,
  title={Projection-Manifold Regularized Latent Diffusion for Robust General Image Fusion},
  author={Cao, Lei and Zhang, Hao and Li, Chunyu and Ma, Jiayi},
  journal={Advances in Neural Information Processing Systems},
  volume={38},
  pages={83972--84003},
  year={2026}
}

📬 Contact

If you have any questions or discussions, please send me an email:

whu.caolei@whu.edu.cn