---
title: LAE-DINO
canonical_url: "https://www.modelscope.cn/studios/ML4Sustain/LAE-DINO"
md_url: "https://www.modelscope.cn/studios/ML4Sustain/LAE-DINO.md"
repository: ML4Sustain/LAE-DINO
chinese_name: "定位在地球上的一切"
last_updated: 2026-06-07
sdk_type: gradio
sdk_version: 6.8.0
downloads: 0
stars: 3
---

# LAE-DINO

> LAE-DINO - ML4Sustain 在 ModelScope 创建的在线 Demo。Locate Anything on Earth

ML4Sustain/LAE-DINO 是 ModelScope 魔搭社区上的在线可交互 Demo（创空间），基于 gradio 6.8.0 构建，中文名为「定位在地球上的一切」。

- **Repository**: ML4Sustain/LAE-DINO
- **SDK**: gradio
- **SDK version**: 6.8.0
- **Downloads**: 0
- **Stars**: 3
- **Last updated**: 2026-06-07

Source: https://www.modelscope.cn/studios/ML4Sustain/LAE-DINO

---

# LAE-DINO Gradio App

This Gradio application provides an easy-to-use interface for deploying **LAE-DINO**, an open-vocabulary object detector for remote sensing imagery. Users can upload an image, specify target object categories, adjust detection thresholds, and run inference through the official LAE-DINO demo pipeline.

## Included Components

This repository contains the source code and configuration files required for LAE-DINO inference. Model weights are automatically downloaded from the ModelScope Model Hub when the application starts.

Included resources:

* Official inference script: `mmdetection_lae/demo/image_demo.py`
* Default configuration:
  `mmdetection_lae/configs/lae_dino/lae_dino_swin-t_pretrain_LAE-1M.py`
* LAE-DINO checkpoint:
  `ML4Sustain/LAE-DINO/checkpoints/lae_dino_swint_lae1m-28ca3a15.pth`
* BERT backbone:
  `google-bert/bert-base-uncased`

Weights are downloaded using `modelscope.snapshot_download`. You may also customize the model paths through environment variables:

```bash
export LAE_DINO_WEIGHTS=ML4Sustain/LAE-DINO
export LAE_DINO_CHECKPOINT_FILE=checkpoints/lae_dino_swint_lae1m-28ca3a15.pth
export LAE_DINO_BERT_MODEL=google-bert/bert-base-uncased
export LAE_DINO_DEVICE=cuda:0
```

## Environment Setup

LAE-DINO depends on OpenMMLab and CUDA-enabled PyTorch.

The official recommended environment is:

* Python 3.8
* CUDA 11.3

Installation:

```bash
cd mmdetection_lae

pip install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 \
-f https://download.pytorch.org/whl/cu113/torch_stable.html

pip install -U openmim

mim install mmengine
mim install "mmcv>=2.0.0"

pip install -v -e .
pip install -r requirements/multimodal.txt

pip install emoji ddd-dataset
pip install git+https://github.com/lvis-dataset/lvis-api.git

cd ..
```

## Run Locally

```bash
pip install -r requirements.txt
python app.py
```

If required dependencies such as `torch`, `mmcv`, or `mmengine` are missing, the error messages from the official inference script will be displayed in the application log panel.

## Citation

If you find LAE-DINO useful in your research, please cite:

**Locate Anything on Earth: Advancing Open-Vocabulary Object Detection for Remote Sensing Community**

Jiancheng Pan, Yanxing Liu, Yuqian Fu, Muyuan Ma, Jiaohao Li, Danda Pani Paudel, Luc Van Gool, and Xiaomeng Huang.

ArXiv: 2408.09110

https://arxiv.org/abs/2408.09110

### BibTeX

```bibtex
@article{pan2024locate,
  title={Locate Anything on Earth: Advancing Open-Vocabulary Object Detection for Remote Sensing Community},
  author={Pan, Jiancheng and Liu, Yanxing and Fu, Yuqian and Ma, Muyuan and Li, Jiaohao and Paudel, Danda Pani and Van Gool, Luc and Huang, Xiaomeng},
  journal={arXiv preprint arXiv:2408.09110},
  year={2024}
}
```