README.md

---
base_model:
- Qwen/Qwen2.5-VL-7B-Instruct
datasets:
- earth-insights/EarthReason
library_name: transformers
pipeline_tag: image-segmentation
license: apache-2.0
---

# Bridging Semantics and Geometry: A Decoupled LVLM–SAM Framework for Reasoning Segmentation in Remote Sensing

This repository contains the 7B model of **Think2Seg-RS**, a decoupled framework for reasoning segmentation in remote sensing (RS) imagery. 

The model was introduced in the paper [Bridging Semantics and Geometry: A Decoupled LVLM-SAM Framework for Reasoning Segmentation in Remote Sensing](https://huggingface.co/papers/2512.19302).

## Overview

Think2Seg-RS decouples high-level semantic reasoning from low-level geometric execution. It trains an LVLM prompter (based on Qwen-2.5-VL) to control a frozen Segment Anything Model (SAM2) via structured geometric prompts. Through a mask-only reinforcement learning objective, the LVLM learns to translate abstract semantic reasoning into spatially grounded actions, achieving state-of-the-art performance on the EarthReason dataset.

## Resources

- **Paper:** [arXiv:2512.19302](https://huggingface.co/papers/2512.19302)
- **Code:** [GitHub - Think2Seg-RS](https://github.com/Ricardo-XZ/Think2Seg-RS)
- **Dataset:** [EarthReason](https://huggingface.co/datasets/earth-insights/EarthReason)

## Citation

If you find this work helpful for your research, please cite:

```bibtex
@article{think2seg_rs_2025,
  title={Bridging Semantics and Geometry: A Decoupled LVLM-SAM Framework for Reasoning Segmentation in Remote Sensing},
  author={Anonymous},
  journal={arXiv preprint arXiv:2512.19302},
  year={2025}
}
```