PD12M (Public Domain 12M)
PD12M is the largest image-text dataset drawn exclusively from public domain and CC0 material, which means the source images carry no copyright restrictions. At 12.4M image-caption pairs it matches the scale of widely used web-scraped datasets, so you can train comparable models without the licensing risk that comes with scraping images under unknown terms.
The images are hosted on dedicated cloud storage separate from their original websites. This addresses link rot, the slow decay of web-scraped datasets as the pages they point to go offline, which leaves many older image-text datasets full of dead links over time.
A smaller 3.3M item subset, PD3M, is available for teams that want the same clean licensing at a more manageable size.
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.