Kraken segmentation model for two-column prints

Segmentation model for two-column prints, namely Wiener Blätter (1851), Der Orient (1850-1851), Allgemeine Zeitung des Judenthums (1848) and Biographisches Lexikon böhmischen Länder (I/1). Training data are specified in models.json, trained region and line types are specified in documentation.pdf.