Christof Rolker · PyLaia · Published October 2, 2026
Beneventana 1.7
Text Recognition
Description
The model was trained by finetuning a model trained mainly from Vat. lat. 5845, a South-Italian codex written in the tenth century in Beneventana. For more information on the codex, see https://data.mgh.de/databases/clavis/wiki/index.php/Citt%C3%A0_del_Vaticano,_BAV,_Vat._lat._5845
To my knowledge, this is the first HTR model for the Beneventan script. Transcription follows Catmus guidelines (https://catmus-guidelines.github.io/html/guidelines/en/generalites.html). Ground Truth was exported, reworked using escriptorium and other tools to conform Catmus rules, and reimported to Transkribus.
Try this model
Drag an image here
Select a file...PNG or JPG up to 10 Mb
Wolpi
AI Assistant
By uploading an image, you accept our terms and privacy policy.
Use this modelOpen in Transkribus
Low error rate5.21% CER
Character Error Rate (CER) measures the percentage of characters incorrectly recognised. Lower is better. This model scored 5.21% on its validation set. As a rule of thumb, a CER below 10% is considered good for most handwritten material.
Measured on the model's own validation data. Results on your documents may differ depending on handwriting style, document condition, language, and how closely your material resembles the training data.
Words15,036
Lines3,161
Training Pages55
Model ID643589