The version 1.0 of the dataset contains about 1.700 images and the associated labels for a subset of text regions.