Published March 19, 2026 | Version 1.0
Dataset Open

Astronomy Dataset for Multimodal Models

Authors/Creators

  • 1. ROR icon Arab Academy for Science, Technology, and Maritime Transport

Description

Space Vision Dataset (Multimodal)

Overview

"A multimodal dataset of space imagery paired with descriptive captions for vision-language research."

The Space Vision Dataset contains images of space scenes with aligned text descriptions, including Earth, Mars, telescopes, and galaxies.

Dataset Structure

Field Description
image_id Unique identifier
image Image file path
text Caption describing the image

Example

 
{
"image_id": "001",
"text": "A satellite image of Earth showing the curvature and scattered clouds over the Atlantic Ocean.",
"image": "images/001.jpg"
}

Categories

  • Earth observations

  • Mars exploration

  • Space telescopes

  • Milky Way imagery

  • Mars rover missions

File Organization

 
dataset/
├── dataset.json
└── images/
├── 001.jpg
├── 002.jpg
└── ...
 

Usage

 
from datasets import load_dataset

dataset = load_dataset("AIOmarRehan/space-multimodal-dataset")

Applications

  • Image captioning

  • Vision-language modeling

  • Multimodal retrieval

  • Research and education

License

MIT License

Acknowledgements

"Developed for research and experimentation in multimodal AI systems."

Contact

For questions or contributions, please open an issue.

Files

astronomy_dataset.zip

Files (94.2 MB)

Name Size Download all
md5:31bf82168430d80c8e6692358ddf126e
94.2 MB Preview Download
md5:f6a82b58b6ce9793db7421916c417f45
2.2 kB Preview Download

Additional details

Dates

Collected
2025-10-01
Collected from different images on the internet.

Software

Repository URL
https://huggingface.co/datasets/AIOmarRehan/space-multimodal-dataset
Programming language
Python
Development Status
Active