We apply the TrOCR framework to real-world, historical manuscripts and show that TrOCR per se is a strong model, ideal for transfer learning. TrOCR has been trained on English only, but it can adapt to other languages that use the Latin alphabet fairly easily and with little training material. We compare TrOCR against a SOTA HTR framework (Transkribus) and show that it can beat such systems. This finding is essential since Transkribus performs best when it has access to baseline information, which is not needed at all to fine-tune TrOCR.
Cite
@article{arxiv.2203.11008,
title = {Transformer-based HTR for Historical Documents},
author = {Phillip Benjamin Ströbel and Simon Clematide and Martin Volk and Tobias Hodel},
journal= {arXiv preprint arXiv:2203.11008},
year = {2022}
}
Comments
This is an abstract submitted and accepted at ComHum 2022 in Lausanne. We will be elaborating on these initial findings in the paper that we will submit after the conference