CC3M

The dataset used in this paper is a large-scale web sourced image-text paired dataset.

BibTex: