---
license: mit
task_categories:
  - text-to-image
  - image-to-image
  - mask-generation
  - image-segmentation
language:
  - en
size_categories:
  - 100K<n<1M
source_datasets:
  - OpenEarthMap
  - LoveDA
  - DeepGlobe
  - SAMRS
  - LAE-1M
tags:
  - diffusion-models
  - remote-sensing
  - image-synthesis
  - controlnet
  - earth-observation
  - generative-models
pretty_name: EarthSynth-180K
---


# EarthSynth-180K Dataset

<p align="center">
  <img src="https://jianchengpan.space/EarthSynth-website/assets/EarthSynth-180K.png" alt="EarthSynth-180K" width="600"/>
</p>

**EarthSynth-180K** is a **multi-task, conditional, diffusion-based generative dataset** designed for remote sensing image synthesis and understanding.  
It was introduced in the paper *"EarthSynth: Generating Informative Earth Observation with Diffusion Models"* (arXiv 2025).  

This dataset supports **text-to-image generation**, **mask-conditioned synthesis**, and **multi-category augmentation** for Earth observation research.

---

## Dataset Details

### Dataset Description

- **Curated by:** Jiancheng Pan, Shiye Lei, Yuqian Fu, Jiahao Li, Yanxing Liu, Yuze Sun, Xiao He, Long Peng, Xiaomeng Huang, Bo Zhao  
- **Funded by:** [Not specified]  
- **Shared by:** EarthSynth Team  
- **Language(s):** English (for prompts)  
- **License:** MIT License  

### Dataset Sources

- **Repository:** [GitHub - EarthSynth](https://github.com/jaychempan/EarthSynth)  
- **Paper:** [ArXiv 2505.12108](https://arxiv.org/abs/2505.12108)  
- **Project Page:** [EarthSynth Website](https://jianchengpan.space/EarthSynth-website/index.html)  
- **Dataset Download:** [HuggingFace](https://huggingface.co/datasets/jaychempan/EarthSynth-180K)  

---

## Dataset Structure

| Subset      | # Images  | Annotations        | Format            | Condition Types          |
|-------------|-----------|--------------------|------------------|---------------------------|
| Train        | 180,000   | Masks, Prompts      | PNG + JSONL       | Mask + Text               |
| Validation   | 10,000    | Masks, Prompts      | PNG + JSONL       | Mask + Text               |
| Augmented    | 180,000   | Single-Category     | PNG + JSONL       | Category + Mask + Text     |

- **Masks:** Binary/instance masks for each object category.  
- **Prompts:** Text prompts for conditional generation.  
- **Augmentation:** Single-category augmentation for CF-Comp training strategy.  

---

## Quick Start

```python
from datasets import load_dataset

# Load dataset
dataset = load_dataset("jaychempan/EarthSynth-180K", split="train")

# Access one example
example = dataset[0]
print(example.keys())  # ['image', 'mask', 'prompt']

# Display image
from PIL import Image
import io

img = Image.open(io.BytesIO(example["image"]))
img.show()