Handwritten Line Segmentation on Medieval Charters

869 annotated charter tenors, randomly selected from the Monasterium database. Each image results from an automated layout analysis of the original charter image: it is in most cases a crop of the charter's main text (aka. tenor) or a smaller region of old text that is apart from the main text region. Annotations are provided in two main formats: JSON file ({*.lines.gt.json}): baselines, centerlines, polygon_coordinates, x-height. PageXML file ({*.lines.gt.xml}): baselines and polygon coordinates File {md5_url_collation.tsv} maps file MD5 prefixes with the source charters in the Monasterium database.