THE INFLUENCE OF GREYSCALE IN GALAXIES CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORKS

Authors

  • J. P. HOLANDA Physics Department, Federal University of Piaui
  • R. L. S. SANTOS Computer Department, Federal University of Piaui

Keywords:

Galaxies, Convolutional Neural Network, Greyscale

Abstract

Galaxies are distributed in various shapes in the universe, in which their arrangements are primary sources for knowledge about their formation and evolution. The classification of these groups of celestial bodies, gas and dust is defined as morphological, which is based on physical and visual aspects. In the last decades, surveys carried out by telescopes have resulted in large sets of images, making it impossible in the human aid for classification, thus enabling the use of methods of artificial intelligence. In this work we used images belonging to the Galaxy10 DECals, in which 400 were separated in two classes: Edge-on and Spiral, submitted to the pre-processing in two greyscale algorithms, known as Gleam and Luma. For the galaxies identification, a convolutional neural network (CNN) was used, based on LeNet-5 architecture, aiming to achieve high levels of accuracy. The results found indicated differences in the metrics of the trained CNN, showing that the Luma algorithm used in the images offers a higher assertiveness than the Gleam, verified in the decrease of errors in certain redshift intervals.

Downloads

Published

2022-12-10

How to Cite

HOLANDA, J. P., & SANTOS, R. L. S. (2022). THE INFLUENCE OF GREYSCALE IN GALAXIES CLASSIFICATION USING CONVOLUTIONAL NEURAL NETWORKS. Journal of Production and Automation (JPAUT) ISSN 2595-9573, 5(2), 38–42. Retrieved from https://jpaut.com.br/index.php/jpaut/article/view/11