Detecting interstellar bubbles in images using CNNs
Ανίχνευση κοσμικών φυσαλίδων σε εικόνες με χρήση ΣΝΔ
View/
Keywords
ISM Bubbles ; Convolutional Neural Network ; Deep learning ; Image segmentationAbstract
Early-type stars create stellar-wind bubbles often with irregular shapes, causing their detection to be both time-consuming and prone to human biases. In this dissertation, we present a program to create and train a Convolutional Neural Network (CNN) to detect these stellar objects which are termed Interstellar Bubbles by astronomers, using image detection. These bubbles have been areas of interest for a while as they have been observed to be a hosting environment for star formation and because they are visually distinct. Recent advances in deep learning, particularly CNNs, have allowed an automated approach to detect these structures with improved efficiency and accuracy. Here we present a CNN model with a Residual U-Net architecture, trained on the Milky Way Project data from Spitzer telescope, that identifies bubbles with 91% accuracy and nearly 1% false positive rate and outputs their location and shape parameters.
Department
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα ΠληροφορικήςNumber of pages
38Language
EnglishCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα