LAION-2B

The dataset used in the paper is LAION-2B, which is a large-scale image-text dataset. The authors fine-tune a pre-trained diffusion model with a subset of LAION-2B with 10k randomly selected samples.

BibTex: