dh-unibe/trocr-kurrent

TrOCR Kurrent-Model 19th century

Handwritten Text Recognition model for 19th century German.

Part of the developments at the Digital Humanities@University of Bern. Developed by Jonas Widmer and Tobias Hodel in conjunction with researchers and institutions mentioned below.

Base model: microsoft/trocr-base-handwritten

Train Lines: 292'997 Eval Lines: 7'513 Test Lines: 15'817

Epochs: 19.66 / 20 Eval CER: 0.02827 Test CER: 0.02655

Finetuned on Kurrent-dataset, containing:

• —Material from the State Archives of Zurich ("Regierungsratsprotokolle"), provided by the State Archives of Zurich
• —Lecture notes of Humboldt Lectures, provided by the Berlin-Brandenburgian Academy of Sciences
• —Diary of Eugen Huber, provided by the University of Zurich
• —Handwriting and Copies by and of Gottfried Semper (provided by the respective research project at ETH Zürich and USI Mendrisio)
• —Konzilsprotokolle, University of Greifswald (19th century)
• —as well as many other smaller collections/examples

The model has not been extensively tested. Potential biases are still to be identified.

Input: one line, not a page

This is a VisionEncoderDecoderModel for line images: it reads one cropped text line per call, which is also what the CER above was measured on. A whole page is not an input it has — pass one and it returns a fragment of it.

Segment the page first (kraken, eScriptorium, Transkribus, or any line segmenter) and pass the crops one at a time. In the DH Bern serving stack the model is registered level: line and the gateway segments with kraken before calling it; the reasoning and the measurements behind that rule are in thodel/serving-atr-inference#165.