- “The RASAM dataset was developed by the Research Consortium Middle-East and Muslim Worlds (GIS MOMM), DISTAM, Calfa, and the BULAC Library between 2021 and 2023. It comprises a diverse collection of Maghrebi Arabic manuscripts from the BULAC Library, featuring a wide variety of handwriting styles, layouts, states of preservation, and other characteristics representative of Arabic Maghrebi manuscript production. The dataset is specifically designed to address the challenges of Handwritten Text Recognition (HTR) for this non-Latin, under-resourced script. In total, the dataset includes 450 images paired with corresponding ground truth data. It covers a representative part of the handwritten production in Arabic Maghrebi scripts and is designed for training generic Arabic handwritten text recognition (HTR) models.”