Abstract

Contributed Talk - Splinter EScience

Thursday, 12 September 2024, 14:25   (S13)

Rotational invariance for galaxy morphology classification

Renuka Velu, Kai Polsterer
Heidelberg Institute for Theoretical Studies

The study of galaxy morphology is essential for understanding the formation and evolution of galaxies, as well as the underlying physical processes. We have a large number of images of galaxies from various surveys and telescopes, which provide sufficient information to carry out the morphology classification. Traditionally, this has relied on visual inspection by experts, which is time-consuming and impractical for large datasets. With advancements in deep neural networks and the availability of extensive datasets, automating this process has become feasible. Building upon the work by Dieleman (2014), the winner of the Galaxy Zoo challenge, we aim to experiment and understand how rotational invariance is achieved in galaxy morphology classification, using publicly available data from the Galaxy Zoo project. This method will facilitate robust classification of galaxy morphology from the vast data provided by next-generation telescopes.