Discusses the development of a deep learning algorithm to detect and segment edge-on galaxies in astronomical images. Edge-on galaxies are of great interest in galactic astrophysics due to their unique orientation, which allows for the study of various galactic phenomena.
The process begins with the selection of a dataset of edge-on spiral galaxies from the Galaxy Zoo project and the corresponding images from the Sloan Digital Sky Survey (SDSS). Approximately 16,000 galaxies were selected and used to train the YOLOv5 algorithm for detection purposes. To isolate galaxies from their backgrounds, the SCSS-Net neural network was used to generate segmentation masks. The algorithm detected 8,000 edge-on galaxies, for which a catalog including their parameters obtained from the SDSS database was compiled.
The algorithm is structured into three main steps:
Detection: Using the YOLOv5 model, detection boxes of edge-on galaxies are obtained.
Segmentation: Extraction of cutouts of detected galaxies using a U-Net-based architecture.
Post-processing: Generation of segmentation masks and extraction of galaxy parameters for further analysis.
Initial data were selected from Galaxy Zoo 2, following specific criteria to ensure that the selected galaxies were spiral and edge-on. The selected images were uploaded to the Zooniverse platform for data annotation by volunteers, which allowed for accurate and verifiable annotations. This collaboration enabled the preparation of the data for the development of the machine learning algorithm.
The paper highlights that most detected galaxies have redshifts between 0.02 and 0.10, with low b/a values (ratio of minor to major axes), and are mostly red, which is consistent with the expected characteristics of edge-on galaxies.
The methodology demonstrates how combining detection and segmentation algorithms can automate the identification and preparation of galaxies for scientific studies. The integrated approach from raw data to usable scientific results is a significant innovation in galactic astrophysics.
Concludes that the algorithm is not only capable of detecting and segmenting edge-on galaxies with high precision but can also be applied to data from future astronomical surveys, extending its utility and application in future galactic studies.
This work represents a significant advance in the application of artificial intelligence techniques for analyzing large volumes of astronomical data, facilitating the identification and study of edge-on galaxies, which are crucial for better understanding the structure and dynamics of galaxies.
Nota: Text generated with the help of AI.
Top comments (0)