Galaxy Shape Categorization Using Convolutional Neural Network Approach

To understand the construction and evolution, galaxy morphological classification is one of the important parameters. With the latest telescope, astronomy is encountering a huge amount of data size and complexity. The challenge is to create a trustworthy system for predicting morphology from galaxy images. We present the Convolutional Neural Network (CNN) model for morphological classification of galaxies in 10 classes (i.e., Face-on, Smooth, in-between, Cigar shaped, Rounded Bulge, Boxy Bulge, No Bulge, Tight Spiral, Medium Spiral, Loose Spiral) with a sample size of 21,785 galaxy images from Galaxy Zoo 2 dataset which were taken from Solan Digital Sky Survey (SDSS) telescope. The proposed CNN model performed well and give an accuracy of 89.86% on training and 85% in testing.