CosmoCLIP: Generalizing Large Vision-Language Models for Astronomical Imaging

TL;DR

CosmoCLIP is a specialized astronomical image-text contrastive learning framework that fine-tunes a pre-trained CLIP model using SpaceNet and BLIP captions, achieving superior zero-shot performance in astronomical tasks.

Contribution

We introduce CosmoCLIP, a novel astronomical contrastive learning framework that leverages SpaceNet and BLIP captions to enhance generalization of vision-language models in astronomy.

Findings

Outperforms CLIP in zero-shot classification

Achieves superior image-text retrieval accuracy

Demonstrates strong generalization across tasks

Abstract

Existing vision-text contrastive learning models enhance representation transferability and support zero-shot prediction by matching paired image and caption embeddings while pushing unrelated pairs apart. However, astronomical image-label datasets are significantly smaller compared to general image and label datasets available from the internet. We introduce CosmoCLIP, an astronomical image-text contrastive learning framework precisely fine-tuned on the pre-trained CLIP model using SpaceNet and BLIP-based captions. SpaceNet, attained via FLARE, constitutes ~13k optimally distributed images, while BLIP acts as a rich knowledge extractor. The rich semantics derived from this SpaceNet and BLIP descriptions, when learned contrastively, enable CosmoCLIP to achieve superior generalization across various in-domain and out-of-domain tasks. Our results demonstrate that CosmoCLIP is a straightforward yet powerful framework, significantly outperforming CLIP in zero-shot classification and image-text retrieval tasks.