Conceptual 12M is a dataset containing (image-URL, caption) pairs collected for vision-and-language pre-training.
No resources for this project.