Detect Nebulae and Galaxies Easier With This Astronomical Image Dataset

Want to train a better deep learning model for astronomical object detection? The Augmented-DeepSpaceYolo dataset might be your secret weapon. This enhanced dataset builds upon the original DeepSpaceYoloDataset and offers a robust resource for training effective Deep Sky Object (DSO) detection models. Learn how it can help you.

What is the Augmented-DeepSpaceYolo Dataset?

This dataset is designed to detect deep sky objects like nebulae, galaxies, and globular clusters in astronomical images. It's an augmented version of the original DeepSpaceYoloDataset, created for the study "Deep Sky Object Detection in Astronomical Imagery Using YOLO Models: A Comparative Assessment."

Why Use Augmented Data for Deep Sky Object Detection?

The Augmented-DeepSpaceYolo dataset dramatically increases the number of images available for training, from 4,696 to 8,421. This increase is achieved through carefully selected data augmentation techniques designed specifically to improve model performance in real-world astronomical imaging scenarios. By simulating variations in image quality, you train your models to be more robust and accurate in identifying DSOs.

What Data Augmentation Techniques Were Used?

To reflect real-world challenges in capturing astronomical images, the following transformations were applied to the original dataset:

• Cropping & Zooming: Simulates variations in focal length and viewing perspectives (0% minimum zoom, 20% maximum zoom).
• Rotations: Accounts for telescope movement and the Earth's rotation (between -15° and +15°).
• Brightness Adjustments: Replicates fluctuating light conditions and atmospheric interference (-15% to +15%).
• Blurring: Mimics slight defocusing and turbulence effects (up to 2.5 pixels).
• Noise Addition: Replicates sensor noise and other random artifacts affecting image clarity (up to 0.15% of pixels).

Each technique addresses specific challenges that can interfere with accurate deep sky object detection.

How is the Augmented-DeepSpaceYolo Dataset Organized?

The dataset includes a default train/validation split, which was used in the associated research paper. Images are provided in JPG format with a resolution of 608x608 pixels. Crucially, labels are provided in the YOLO format, ensuring compatibility with popular object detection frameworks. This makes integrating Augmented-DeepSpaceYolo into your projects straightforward.

Quick List of the Benefits of Using This Dataset

• Increased Data Volume: More images for better model training.
• Realistic Augmentations: Mimics real-world astronomical image imperfections.
• Ready-to-Use Format: YOLO-formatted labels for easy integration.
• Improved Model Robustness: Enhanced ability to handle image variations.
• Better Deep Sky Object Detection: Detect nebulae and galaxies with higher accuracy.

How Can I Download the Augmented-DeepSpaceYolo Dataset?

The dataset is readily available for download via [Google Drive](insert Google Drive link here). Start experimenting and improving your DSO detection models today.

How to Contribute

The open source community thrives on contributions! If you have ideas to improve the Augmented-DeepSpaceYolo dataset, consider contributing:

1. Fork the repository.
2. Create a new feature branch (git checkout -b feature/NewFeature).
3. Commit your changes (git commit -m 'Add some NewFeature').
4. Push to the branch (git push origin feature/NewFeature).
5. Open a pull request.

Don't forget to star the project to show your support!

If the Augmented-DeepSpaceYolo dataset proves helpful in your research, please cite the following article:

@article{ramos_rivas-echeverría_2025,
title={Deep sky object detection in astronomical imagery using YOLO models: a comparative assessment},
DOI={https://doi.org/10.1007/s00521-025-11223-4},
journal={Neural Computing and Applications},
publisher={Springer Science and Business Media LLC},
author={Ramos, Leo Thomas and Rivas-Echeverría, Francklin},
year={2025},
month={Apr}}