WIT (Wikipedia-based Image Text)
WIT draws on Wikipedia to connect images with the text around them. For each image it captures the caption, the reference description, and passages from the article the image sits in, so a model gets far more context than a one-line caption provides.
Its defining feature is language coverage. WIT spans well over 100 languages, which makes it one of the few large image-text datasets suitable for training and evaluating multilingual multimodal systems rather than English-only ones.
The dataset is released by Google Research under CC BY-SA 4.0, a share-alike licence, meaning any derivative database you distribute has to carry the same terms. The underlying images come from Wikimedia Commons under their own varied licences, so the picture licences can differ from the dataset licence.
Related sources
Conceptual Captions (CC3M / CC12M)
A cleaned image alt-text dataset for automatic image captioning, produced by Google Research. Alt-text pulled from the web is filtered and generalised (hypernymed), replacing specific names with broader categories. Available in 3.3M (CC3M) and 12M (CC12M) variants.
DataComp
A benchmark and dataset collection for training multimodal models. It provides a shared pool of image-text candidates and an evaluation framework so teams can test different data-curation strategies against a common yardstick.
LAION-5B
An image-text dataset of over 5.8B examples, built by filtering Common Crawl with a CLIP model that scores how well an image matches a caption. Includes 2.32B English pairs, 2.26B multilingual pairs, and 1.27B not tied to any particular language. Provides URLs, not the images themselves.
OBELICS
A web-scale dataset of 141M multimodal English web documents that interleave text and images, containing 353M images and 115 billion tokens. Unlike image-caption pair datasets, it keeps whole documents with text and images in their original order. Used to train the IDEFICS models.