This dataset contains 500,000 synthetic line images generated for the Silver stage of the MEDUSA training curriculum. It was produced at the École nationale des chartes – PSL as part of the MEDUSA project for multilingual medieval handwritten text recognition (HTR).

Motivation

Many medieval languages — particularly Germanic, Celtic, and Slavic ones — are underrepresented in existing image–text HTR corpora. Text-only (Silver) resources, however… See the full description on the dataset page: https://huggingface.co/datasets/ENC-PSL/MEDUSA_Synthetic_Lines.