Machine Learning for Morphological Galaxy Classification

The "Machine learning for morphological galaxy classification" is a repository for classifying Galaxy Zoo 2 (GZ2) images into (1) Galaxy and Non-Galaxy, and (2) Galaxy in Spiral, Elliptical, and Odd objects using the five state-of-the-art machine learning models.

Overview

We employed five different classification models, including:

1. Support Vector Machine (SVM) with Zernike moments (ZMs)
2. 1D-Convolutional Neural Network (1D-CNN) with ZMs
3. 2D-CNN with Vision Transformer (ViT) and original images
4. ResNet50 with ViT and original images
5. VGG16 with ViT and original images

The SVM and 1D-CNN models utilized Zernike moments (ZMs) extracted from the images, while the 2D-CNN, ResNet5, and VGG16 with Vision Transformer (ViT) models were designed based on the original images.

For more details on the algorithms, please refer to our paper: H. Ghaderi, N. Alipour, and H. Safari.

Repository Structure

This repository includes two main Jupyter notebooks:

• Galaxy-Non-Galaxy Classification: galaxy_nongalaxy_classifiers.ipynb
• Galaxy Classification: galaxy_classifiers.ipynb

Data

Please download the Data files from this link that includes two categories:

1. galaxy-nongalaxy
2. galaxy

Each category contains two folders:

• images: This folder includes the original images for galaxy_nongalaxy and cropped images for galaxy classifiers.
• ZMs: This folder contains Zernike Moments (ZMs) data sets in CSV file format.

Authors

Hamed Ghaderi, Nasibe Alipour, Hossein Safari

License

MIT