A novel CNN architecture for the classification of galaxies

Vaid, Vibha (2022) A novel CNN architecture for the classification of galaxies. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (3MB) | Preview
Preview	PDF (Configuration manual) Download (864kB) | Preview

Abstract

The universe and its galaxies have always been an interesting subject to scientists and astronomers. It becomes essential for an astronomer or scientist to classify a galaxy or a deep space object. In this research, a novel Convolution neural network architecture is being developed to classify the morphology of the galaxies. To perform classification the dataset for this project is collected from Kaggle which was a part of the galaxy zoo project. To develop the novel CNN architecture, literature has been reviewed and important aspects of the paper are extracted. A concatenated CNN has also been developed in order to combine the 2 neural networks and produce better results. The CNN and concatenated CNN are compared with other existing models of transfer learning like vgg19, densenet121, and inceptionv3. The main goal of this research has been to develop a novel convolutional neural network and concatenated CNN which improve the accuracy of the classification of galaxies. The CNN architecture is developed from scratch by concatenating two different networks. The research follows the knowledge discovery in database (KDD) approach for classification. The models are tested on the accuracy, precision, and recall. It can be seen that inceptionv3 gives the most accurate results in comparison to other neural networks used. It was observed that the concatenated model did not converge and performed poorly in comparison to the other networks

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Image classification; vgg19; densenet121; inceptionv3; CNN; Concatenated CNN
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QB Astronomy
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	14 Mar 2023 12:16
Last Modified:	14 Mar 2023 12:16
URI:	https://norma.ncirl.ie/id/eprint/6330

Actions (login required)

View Item