Crossing the Bifrost: Towards an open access FAIR HTR model for Old Norse manuscripts

This data deposit contains the dataset behind the DH2025 paper, ‘Crossing the Bifrost: Towards an open access FAIR HTR model for Old Norse manuscripts’ by Katarzyna Anna Kapitan & Chahan Vidal-Gorène. The zip-file contains two folders. 1. The folder bifrost_01_training_data contains JPGS and ALTO-XMLs for selected leaves from two Old Norse manuscripts which were used for fine-tuning the CATMuS Medieval model released by Pinche, A., Clérice, et al. (2025). CATMuS Medieval. Zenodo. https://doi.org/10.5281/zenodo.15030337. This dataset is a very slightly revised copy of the dataset released as Kapitan, K. A., Latham, C., Pinche, A., Verri, G., & Yavuz, N. K. (2024). TranscriboQuest24 Icelandic [Data set]. Zenodo. https://doi.org/10.5281/zenodo.13757377. 2. The folder bifrost_01_test_data contains JPGS and ALTO-XMLs for selected leaves from four Old Norse manuscripts which were used for testing the finetuned model’s performance. This is a dedicated dataset developed for the DH2025. The fine-tunned model has been released separately as bifrost_01 for those interested in testing it.