Galaxy morphology classification using deep learning: a compact survey

Galaxy morphology classification is a key task in extragalactic astronomy, as structural and dynamical properties of galaxies are closely tied to their formation history, stellar populations, and environment. Next-generation surveys such as the Sloan Digital Sky Survey (SDSS), the Vera C. Rubin Observatory’s Legacy Survey of Space and Time (LSST), and the Euclid mission are producing petabyte-scale imaging data, making manual classification impractical. In view of this, deep learning (DL)-based automated methods provide a scalable and efficient alternative. This survey reviews DL methods for galaxy morphology classification over the past five years focusing on convolutional neural networks (CNNs), vision transformers (ViTs), graph neural networks (GNNs), and hybrid architectures. A taxonomy is introduced to categorize models by architecture, metadata integration, and interpretability, supported by comparative analysis of datasets, generalization ability, and efficiency. This survey highlights several key gaps, including limited cross-survey robustness, underuse of astrophysical metadata (e.g., redshift, Sérsic index, velocity dispersion), insufficient interpretability of learned features such as spiral arms or bulge-to-disk ratios, and high computational cost of advanced architectures. To address these gaps, promising directions involve lightweight and metadata-aware models, multimodal frameworks integrating spectroscopy and imaging, self-supervised and physics-informed methods, and approaches incorporating temporal and spatial evolution. By aligning machine learning progress with astrophysical insight, future models can achieve accuracy, scalability, and scientific interpretability, ultimately advancing our understanding of galaxy structure and evolution.