View Single Post
Old 02-16-2022, 04:40 PM   #11
Jannu
Connoisseur
Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.Jannu ought to be getting tired of karma fortunes by now.
 
Posts: 58
Karma: 718318
Join Date: Jan 2021
Device: Kindle Paperwhite, iPad Air, Fire HD 8 + LineageOS
trainedata Dateien für tesseract Fraktur der UB-Mannheim

Um den thread mal wieder auf einen aktuellen Stand zu bringen.

Die Universitätsbibliothek Mannheim hat einiges an Aufwand in das Training von traineddata Dateien für tesseract Fraktur gesteckt (bis 13000 Stunden CPU-Zeit und 450000 Iterationen für eine traineddata).

Der passende Artikel dazu mit den Links auf die traineddatas findet sich hier:

https://github.com/tesseract-ocr/tes...iki/GT4HistOCR

Ich hab bei Fraktur-Schriften gute Erfahrungen mit folgender traineddata (von 2021) gemacht:

https://ub-backup.bib.uni-mannheim.d...05.traineddata
Jannu is offline   Reply With Quote