How to finetune DINOv2 for astronmy?
Abstract
This study evaluates the performance of existing visual foundation models, based on ViT, SwinV2, BEiT or DINOv2, for astronomical applications, particularly galaxy morphological classification. We explore different fine-tuning strategies to specialize DINOv2 for astronomy and report results on the Galaxy10 DECaLS dataset.
Type
Publication
GRETSI 2025
Authors
AI for astronomy & astrophysics
PhD student at IRISA, Université Bretagne Sud (expected 2026), and lecturer in computer science,
working on foundation models for astronomy and astrophysics, with a broader interest
in deep learning applied to scientific data.
Also contributing to the UniverseTBD collaboration on vision–language models for astronomy.