Dark

Light

Alto-44

OCR scores

Historical newspapers OCR ground-truth

A dataset consisting of 2000 pages historical newspaper groundtruth, OCR and images.

An example of an image and caption extracted from the front page of the January 27th 1951 issue of the De Nieuwsgier.

KBK-1M

The KBK-1M Dataset is a collection of 1,603,396 images and accompanying captions from 1922 – 1994