---

📢 Latest Updates

• 2024.12.10 Release huggingface demo
• 2024.12.8 Release demo code

---

🔨 Installation

1. Clone this repo:

git clone git@github.com:xuliu-cyber/RSUniVLM.git
cd RSUniVLM

2. Install required Python packages

conda create -n rsunivlm python=3.11
conda activate rsunivlm
pip install --upgrade pip  # Enable PEP 660 support.
pip install -e ".[train]"
pip install gradio

🖼️ Demo

1. Download our available checkpoints.
2. Change the model_path to your checkpoint path.
3. Execute the code to run an interactive demo on your own machine:

python app.py

🎉 Acknowledgement

• Code & Model: We build our project based on the code repo of LLaVA-NeXT.

🧩 Citation

If you find this work useful, please cite:

@article{LIU2026113717,
title = {RSUniVLM: A unified vision-language model for remote sensing via Granularity-oriented MoE},
journal = {Pattern Recognition},
volume = {179},
pages = {113717},
year = {2026},
issn = {0031-3203},
doi = {https://doi.org/10.1016/j.patcog.2026.113717},
url = {https://www.sciencedirect.com/science/article/pii/S0031320326006825},
author = {Xu Liu and Zhouhui Lian},
keywords = {Remote sensing, Vision language model},
}