Abstract Details

Name: Sheelu Abraham
Affiliation: IUCAA, Pune, India
Conference ID: ASI2017_470
Title : Automated Detection of Barred Galaxies using Convolutional Neural Network
Authors and Co-Authors : Sheelu Abraham, IUCAA, Pune, India-411007; Arun Kumar Aniyan, SKA South Africa, Pinelands,Cape Town 7405, South Africa; Ajit K. Kembhavi, IUCAA, Pune, India - 411007
Abstract Type : Oral
Abstract Category : Extragalactic astronomy
Abstract : Bars are one of the prominent features in most of the disc galaxies. It is observed that a significant fraction of disc galaxies are barred in the near universe and bars play a major role in the secular evolution of disc galaxies. So it is important to detect bars and proper identification of barred galaxies enables us to understand the evolution of disc galaxies. We present a method for detecting bars in galaxies using deep learning. Deep learning methods enable automatic feature learning and extraction directly from the images rather than handcrafted features by humans thereby enabling optimised learning and high accuracy classification. Such methods have significant importance with future observatories where Petabytes of data is generated on a daily basis. We trained a deep convolutional neural network using SDSS DR12 images. Our classifier is able to distinguish barred and unbarred galaxies with ~93% accuracy. We used our classifier to detect bars from Meert et al. 2015 catalogue where they have done bulge-disc decomposition of ~7 x 10^5 galaxies from SDSS DR7.