International Conference on Computer and Knowledge Engineering

Home / 15th International Conference on Computer and Knowledge Engineering

Learning to Classify Messier Astronomical Objects with Limited Data: A Few-Shot Learning Approach

Authors :

AMIRREZA ROUHBAKHSHMEGHRAZI¹ Shayan Nalbandian² Ghazal Alizadeh³ Sheida Shadman⁴ Shuyuan Yang⁵ Bo Li⁶

1- School of Electronics and Information, Northwestern Polytechnical University, Xi'an, China 2- School of Software Engineering, Northwestern Polytechnical University, Xi'an, China 3- School of Aeronautics, Northwestern Polytechnical University, Xi’an, China 4- School of Software Engineering, Northwestern Polytechnical University, Xi'an, China 5- School of Artificial Intelligence, Xidian University, Xi’an, China 6- School of Electronics and Information, Northwestern Polytechnical University, Xi'an, China

Keywords :

MAML،RelationNet،Meta lerning،Astronomy،FSL

Abstract :

Deep learning has achieved remarkable success in image classification; however, its performance heavily relies on large labeled datasets, which are often unavailable in specialized domains like astronomy. This poses a significant challenge for classifying rare or diverse celestial phenomena, including Messier objects—well-known astronomical targets cataloged for observational research. To address this data scarcity problem, this study investigates the application of Few-Shot Learning (FSL), specifically Prototypical Networks, to the classification of Messier astronomical images using limited labeled data. The objective of this research is to evaluate the performance of ProtoNet under various few-shot scenarios and benchmark it against other FSL models, including both metric-based (MatchingNet) and gradient-based approaches (MAML, FOMAML, Meta-SGD). Experimental results demonstrate that ProtoNet achieves high classification accuracy (up to 90.5%), especially in 5-shot settings, and converges significantly faster than gradient-based models. Ablation studies further reveal that deeper backbones such as ResNet-50 and ViT-B/16 improve representation quality in low-data regimes. This research represents the first systematic application of FSL to Messier object classification and provides practical insights into efficient learning from limited astronomical data.