Artificial intelligence has advanced significantly in text-to-image generation in recent years. Transforming written descriptions into visual representations has a number of applications, from creating content to helping the blind and telling stories. The researchers have been facing two significant obstacles, which are the lack of high-quality data and copyright issues related to datasets that are scraped from the internet. In recent research, a team of researchers has proposed the idea of building an image dataset under a Creative Commons licence (CC) and using it to train open diffusion models that can outperform Stable Diffusion 2 (SD2).
