Navigation

Development of a model repository and automatic font recognition for OCR-D

Development of a model repository and automatic font recognition for OCR-D

The DFG-approved project is one of 8 projects in the OCR-D initiative, which want to simplify OCR for libraries and archives. The project is an interdisciplinary project in collaboration with the Book Science Lab (Johannes Gutenberg University Mainz), the Digital Humanities Lab (University Leipzig), and the Pattern Recognition Lab (FAU).

Goal: An automatic font recognition can help to select an appropriate pre-trained model. Therefore, not all possible fonts need to actually be recognized since also models trained for a similar script could achieve high quality OCR results. In this project, we want to solve the question how the OCR quality is related to the script similarity and how many OCR models need to be trained.

Link zum Projekt

 

Projektzeitraum: 06/01/2018 12/30/2019

Projektbeteilige:
Gregory Crane, Andreas Maier, Nikolaus Weichselbaumer, Vincent Christlein, Benjamin Kiessling, Christoph Reske