Speaker
Description
Understanding the formation and evolution of galaxies over cosmic time requires a
comprehensive analysis of their morphologies, especially because morphological
features are strongly connected to other galaxy properties such as stellar
populations, environments, and kinematics. However, the growing size of modern
sky surveys has resulted in massive volumes of unclassified galaxies, making
traditional morphological analysis increasingly difficult and time-consuming. In this
work, we employ Convolutional Neural Networks (CNNs) to construct a
morphological catalog from galaxy images obtained by the DECam Local Volume
Exploration Survey (DELVE). To ensure a reliable training sample, we use a subset
of 314,000 galaxies from the Galaxy Zoo DECaLS project (Walmsley et al. 2021),
allowing us to define a robust training set of approximately 98,000 galaxies classified
into four morphological classes: elliptical, lenticular, spiral, and mergers. We
compare our CNN model to widely used architectures from the literature and show
that our model outperforms them in both computational efficiency and classification
accuracy across all morphological types. Our model achieves precision scores of
97%, 98%, 99%, and 92% for elliptical, lenticular, spiral, and merger galaxies,
respectively. Applying this trained model to previously unclassified data from DELVE,
we generate a morphological catalog covering approximately 13 square degrees
down to r-band magnitude 21.5, comprising around 30 million galaxies. The
completion and public release of this catalog will not only enhance our understanding
of galaxy evolution but also provide a valuable resource for the broader astronomical
community.
