Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Novel Sector-Based Algorithm for an Optimized Star-Galaxy Classification

Anumanchi Agastya Sai Ram Likhit, Divyansh Tripathi, Akshay Agarwal

TL;DR

This work tackles scalable star–galaxy classification for mega-surveys by dividing the celestial sphere into $36$ sectors aligned with SDSS-DR18 patterns and training sector-specific inputs with a lightweight CNN. It demonstrates that sector-aware processing yields higher accuracy than state-of-the-art baselines CovNet and MargNet, achieving $0.9525$ accuracy on combined data and a per-epoch time of $25$ s, far faster than baselines. The study also analyzes sector resiliency, including zero-shot evaluations around $0.89$ accuracy on unseen sectors, supporting generalization. The approach offers practical potential for real-time astronomical analysis and provides a scalable framework adaptable to future surveys like LSST by leveraging SDSS-like stripe segmentation.

Abstract

This paper introduces a novel sector-based methodology for star-galaxy classification, leveraging the latest Sloan Digital Sky Survey data (SDSS-DR18). By strategically segmenting the sky into sectors aligned with SDSS observational patterns and employing a dedicated convolutional neural network (CNN), we achieve state-of-the-art performance for star galaxy classification. Our preliminary results demonstrate a promising pathway for efficient and precise astronomical analysis, especially in real-time observational settings.

A Novel Sector-Based Algorithm for an Optimized Star-Galaxy Classification

TL;DR

This work tackles scalable star–galaxy classification for mega-surveys by dividing the celestial sphere into sectors aligned with SDSS-DR18 patterns and training sector-specific inputs with a lightweight CNN. It demonstrates that sector-aware processing yields higher accuracy than state-of-the-art baselines CovNet and MargNet, achieving accuracy on combined data and a per-epoch time of s, far faster than baselines. The study also analyzes sector resiliency, including zero-shot evaluations around accuracy on unseen sectors, supporting generalization. The approach offers practical potential for real-time astronomical analysis and provides a scalable framework adaptable to future surveys like LSST by leveraging SDSS-like stripe segmentation.

Abstract

This paper introduces a novel sector-based methodology for star-galaxy classification, leveraging the latest Sloan Digital Sky Survey data (SDSS-DR18). By strategically segmenting the sky into sectors aligned with SDSS observational patterns and employing a dedicated convolutional neural network (CNN), we achieve state-of-the-art performance for star galaxy classification. Our preliminary results demonstrate a promising pathway for efficient and precise astronomical analysis, especially in real-time observational settings.
Paper Structure (9 sections, 6 figures, 5 tables)

This paper contains 9 sections, 6 figures, 5 tables.

Table of Contents

  1. INTRODUCTION
  2. Proposed Methodology
  3. EXPERIMENTS AND RESULTS
  4. CONCLUSION
  5. Sky Division
  6. Proposed CNN Architecture
  7. Data Splitting for Model Training and Testing
  8. Choosing Sectors: An Overview
  9. Other Sectors and Zero-Shot Resiliency

Figures (6)

  • Figure 1: A sample image reflecting the challenges in identifying star-galaxies in different sectors.
  • Figure 2: 2D Sky Sector Map
  • Figure 3: Data Workflow Diagram
  • Figure 4: SDSS Sky Coverage sdss-dr7
  • Figure 5: A sample image reflecting the challenges in identifying star-galaxies in sectors 7 and 13.
  • ...and 1 more figures