Globular cluster detection in the Galaxy M33 using multi-view learning

Abstract

Understanding globular clusters (GC) is essential for comprehending the evolution of galaxies. However, identifying these clusters in image collections is time-consuming. This necessity has led to the development of an automated method for detecting GCs. Despite GC detection essentially being an object detection problem, modern algorithms struggle to provide reliable results. Inspired by how astronomers identify GCs, we introduce a deep neural network that combines various perspectives of raw image data to enhance input data representation. Subsequently, this network is integrated with the YOLO object detection approach to form the YOLO for Globular Cluster detection (YOLO-GC) model. Experiments conducted on a genuine catalog of GCs in the M33 Galaxy highlight the advantages of our approach for learning representations from multiple perspectives.