Picturing Stories: A Citizen-Science Annotated Dataset of Children’s Book Illustrations

Illustrated children’s books have played a central role in the cultural transmission of narratives, values, and models of childhood for more than three centuries. Yet despite growing interest in computational approaches to cultural heritage and multimodal analysis, large-scale datasets focused on historical children’s book illustrations remain limited. We introduce Picturing Stories, a new citizen-science annotated dataset containing 44,043 illustrations drawn from a collection of 3,388 children’s books published between 1731 and 2011. The dataset combines materials from the Internet Archive’s Children’s Library and the International Children’s Digital Library, spanning historical and contemporary works across 37 languages. Annotations were produced through a 42-week Zooniverse project in which 10,647 volunteers contributed over 1.3 million total annotations, capturing core dimensions of visual narrative, including character presence, setting, objects, actions, emotional intensity, family relations, and caregiving behaviors, alongside a subset of ~27,000 images with two or more captions. This data can be used to study the history of children’s books and benchmark AI for the study of further datasets.