Published October 21, 2022
| Version v1
Dataset
Open
Dataset for the paper "Ground-truth Free Evaluation of HTR on Old French and Latin Medieval Literary Manuscripts"
Authors/Creators
- 1. Centre Jean Mabillon, École nationale des chartes, PSL University
Description
This dataset was used in the context of the article Ground-truth Free Evaluation of HTR on Old French and Latin Medieval Literary Manuscripts, at the Computational Humanities Research 2022 conference.
Predictions.zip contains the XML ALTO for the HTR and Segmentation prediction of around 10 pages of 1900 manuscripts from the Bibliothèque nationale de France.
Varying ground truth contains the original training material for training a classifier for CER classification (see the article).
The ManuscriptsIIIF.csv contains metadata about the manuscripts.
This work was funded by the DIM MAP under the CREMMALab funding.