Tracing Superbubble with Recurrent Convolutional Neural Network (R-CNN) and Galaxy Image Dataset
Organisatsiooni nimi
Scaling and Intelligence Lab (SaiL)
Kokkuvõte
Disc galaxies are filled with gas, dust and stars. A significant fraction of the gas and dust is in a disc in a non-uniform way. Superbubbles are cavities in gas distribution almost devoid of gas - only sparse high-temperature gas remains. These are formed when the supernovae explosions sweep the gas from the galactic region, forming a thin shell. Superbubbles are observed most spectacularly in infrared images (James Webb Space Telescope or PHANGS survey) or radio wavebands (THINGS survey). Having only one telescope in space actively observing in infrared wavelengths makes images on that part of the spectrum hard to come by. We propose a project combining machine learning and image analysis to determine how well these superbubbles are traceable using optical images of galaxies. If the supernovae push the gas from the explosion region, it will either carry the dust with the explosion or destroy it. Optical images trace some imprints of the dust due to its absorption of light. Therefore, we expect the dust distribution and the superbubbles to have some (anti)correlation but not a strong one (https://arxiv.org/pdf/2212.02667, https://academic.oup.com/mnras/article/309/2/332/1029269)
In this project, we use the ray-tracing method along with Recurrent Convolutional Neural Network (R-CNN) to trace Superbubble. The CNN is trained with Optical image dataset of galaxies. The recurrent layers are fed with the image frames to R-CNN and trained to predict the evolution of Superbubble.
In this project, we use the ray-tracing method along with Recurrent Convolutional Neural Network (R-CNN) to trace Superbubble. The CNN is trained with Optical image dataset of galaxies. The recurrent layers are fed with the image frames to R-CNN and trained to predict the evolution of Superbubble.
Lõputöö kaitsmise aasta
2024-2025
Juhendaja
Aikaterini Niovi Triantafyllaki, Rain Kipper, Kallol Roy
Suhtlemiskeel(ed)
inglise keel
Nõuded kandideerijale
Tase
Bakalaureus, Magister
Kandideerimise kontakt
Nimi
KALLOL ROY
Tel
(+372)56051480
E-mail
Kuulutus
Vaata lähemalt