---
title: GeoPix-ft-sior_rsicap
canonical_url: "https://www.modelscope.cn/models/NormanOU/GeoPix-ft-sior_rsicap"
md_url: "https://www.modelscope.cn/models/NormanOU/GeoPix-ft-sior_rsicap.md"
repository: NormanOU/GeoPix-ft-sior_rsicap
last_updated: 2025-02-21
pipeline_tag: multimodal-dialogue
tasks:
  - multimodal-dialogue
parameters: 7.1B
tensor_type:
  - F32
library_name:
  - safetensors
downloads: 293
stars: 0
---

# GeoPix-ft-sior_rsicap

> GeoPix-ft-sior_rsicap - NormanOU 在 ModelScope 开源的模型。GeoPix: Multi-Modal Large Language Model for Pixel-level Image Understanding in Remote Sensing

- **Repository**: NormanOU/GeoPix-ft-sior_rsicap
- **Tasks**: multimodal-dialogue
- **Parameters**: 7.1B
- **Downloads**: 293
- **Stars**: 0
- **Last updated**: 2025-02-21

Source: https://www.modelscope.cn/models/NormanOU/GeoPix-ft-sior_rsicap

---

<div align="center">
<h1>GeoPix: Multi-Modal Large Language Model for Pixel-level Image Understanding in Remote Sensing</h1>

[**Ruizhe Ou**](https://github.com/Norman-Ou)<sup>1</sup> · [**Yuan Hu**](https://github.com/Lavender105)<sup>2,*</sup> · **Fan Zhang**<sup>2</sup> · **Jiaxin Chen**<sup>1</sup> · **Yu Liu**<sup>2,3</sup>

<sup>1</sup>北京邮电大学 · <sup>2</sup>北京大学 · <sup>3</sup>北京大学鄂尔多斯能源研究院 <sup>*</sup>通讯作者

<a href='https://arxiv.org/abs/2401.07519'><img src='https://img.shields.io/badge/arXiv-2501.06828-b31b1b.svg'></a>

</div>

《GeoPix》是遥感领域的新一代像素级多模态大语言模型，支持指引性图像分割等任务。

# 新闻

- [2025.02.20] 我们发布了预训练Checkpoints、推理代码和 Gradio Demo！ [Github](https://github.com/Norman-Ou/GeoPix)
- [2025.01.12] 我们发布了论文[Arxiv](https://arxiv.org/pdf/2501.06828)。

**GeoPix: Multi-Modal Large Language Model for Pixel-level Image Understanding in Remote Sensing** [[Arxiv](https://arxiv.org/pdf/2501.06828)]

# 摘要
In this work, we propose GeoPix, a RS MLLM that extends image understanding capabilities to the pixel level. This is achieved by equipping the MLLM with a mask predictor, which transforms visual features from the vision encoder into masks conditioned on the LLM’s segmentation token embeddings. For more details, please refer to the [paper](https://arxiv.org/pdf/2501.06828).

# 权重下载
您可以直接从 [Huggingface](https://huggingface.co/InstantX/InstantID)、[ModelScope](https://www.modelscope.cn/models/NormanOU/GeoPix-ft-sior_rsicap) 或 [OpenXLab](https://openxlab.org.cn/models/detail/NormanOU/GeoPix-ft-sior_rsicap) 下载模型。您也可以通过 Python 脚本下载模型：

```python
# Huggingface
from huggingface_hub import snapshot_download
snapshot_download(repo_id="Norman-ou/GeoPix-ft-sior_rsicap", local_dir="./pretrained_models")

# ModelScope
from modelscope import snapshot_download
model_dir = snapshot_download("NormanOU/GeoPix-ft-sior_rsicap", local_dir="./pretrained_models")
```

# 引用我们📑
```
@misc{ou2025geopixmultimodallargelanguage,
      title={GeoPix: Multi-Modal Large Language Model for Pixel-level Image Understanding in Remote Sensing}, 
      author={Ruizhe Ou and Yuan Hu and Fan Zhang and Jiaxin Chen and Yu Liu},
      year={2025},
      eprint={2501.06828},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2501.06828}, 
}
```