Colibri

A fine computer vision dataset comprising of illustrations cut out of 19th century children’s and youth books.

More than 53.000 illustrations extracted from 3.400 children’s books published between 1800 and 1925, accompanied by metadata and annotations classifying the content. Images in .jpg format, metadata and annotations in .csv, Public Domain.

The dataset can serve tasks such as object detection in historical images and hand-made illustrations; or analysis of the evolution of the presentation of printed books over time.

Ample documentation is available in the form of a datasheet accompanying the data publication.