HTR model for transcribing Glagolitic sources printed in the 16th century Tübingen-Urach printing style into the Latin script

This model is the result of a collaboration between the University Library Tübingen and the Slavic Department Freiburg. Selections from several early printed Glagolitic sources (16th century) from the Tübingen Urach collection were used as GT. It can be used to transcribe sources printed in the 16th century Tübingen-Urach printing style into the Latin script. The original training set was prepared in Transkribus, whence it was exported and re-used to train this model. It is possible that the export caused some distortions of baselines or line masks, or corruptions in the data, which may have inflated CER (despite manual cleansing prior to Kraken training).