| Name: J Saranya |
| Affiliation: Indian Institute of Science Education and Research, Tirupati |
| Conference ID: ASI2026_308 |
| Title: Identifying warped disk galaxies with machine learning |
| Abstract Type: Poster |
| Abstract Category: Galaxies and Cosmology |
| Author(s) and Co-Author(s) with Affiliation: Saranya J Suguna(Indian Institute of Science Education and Research, Tirupati - 517619, India), Arunima Banerjee(Indian Institute of Science Education and Research, Tirupati - 517619, India) |
| Abstract: Galaxies observed edge-on allow the study of the vertical distribution of stars, gas and dust, revealing features such as warps. Disk warps are observed in nearly 50% of nearby spiral galaxies and are thought to arise from several formation and evolutionary processes. However, since warps are faint and occur at large galactocentric radii, their origin and evolution remain poorly understood.
In this work, we present a supervised deep learning approach to identify warped edge-on galaxies from the Pan-STARRS EGIPS survey. The dataset consists of 5,812 galaxies selected with inclination angles close to 90 degrees. The images were aligned with their position angles and warp angles are measured to generate reliable labels. We use Zoobot, which provides pretrained deep learning models designed for galaxy morphology analysis. The network is fine-tuned for warp classification in edge-on galaxies. Grad-CAM is used as an explainable AI technique to generate heatmaps linking model predictions to physically meaningful galaxy features.
Initial training using JPEG images showed limited improvement beyond 76% test accuracy. Ongoing work explores the use of FITS images, which are expected to better preserve faint structures such as warps. The results will be used to study the intrinsic and environmental properties of warped galaxies and to better understand their evolutionary status. With upcoming sky surveys and high-resolution telescopes, this approach can be used to detect faint disk warps and assist large-scale galaxy morphology studies.
|
