Published April 26, 2023
| Version 2
Other
Open
Fraktur model trained from enhanced Austrian Newspapers dataset
Description
Recognition model for 19th century German Fraktur texts, trained with ground truth from Austrian Newspapers.
The ground truth which was used for the training is available from Mannheim University Library (see also details on the training process): https://github.com/UB-Mannheim/AustrianNewspapers/
It is based on this dataset:
Günter Mühlberger, & Günter Hackl. (2019). NewsEye / READ OCR training dataset from Austrian Newspapers (19th C.) [Data set]. Zenodo. http://doi.org/10.5281/zenodo.3387369
Notes
improved model based on fixed ground truth
Files
metadata.json
Files
(16.3 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:acb1a9c733248f5a77faf83d12571940
|
16.3 MB | Download |
|
md5:d11eb2ad95d433f01c2003672b7ea29e
|
1.8 kB | Preview Download |