zuphilip/ocropy-french-models

ocropy model for fonts as accuring in old French books

ocropy-french-models

This repository contains

• Trained ocropy models for (old) French with an emphasize to fonts and language as they looked in 16th-19th century.
• The data (image lines together with their transcription) which where used for the training.
• Tools which where used in this process.

Old French Texts

The usual accents in French texts like é, è, ê, á, à, ô, ç are present but old text also use the long-s, which looks very similar to a f. Moreover, the usual ligaturs including the less-often used ct ligatur as well as some special styling of the large letter Q can occur. Here is an example of such a digitized page:

TODOs

• Find a suitable French text
• Find a suitable Type font
• Finish the generation scripts
• Generate the lines
• Start training the model
• Evaluate with real texts
• Improve the model

Acknowledgments

This is free software.

This project bundles other free software, fonts, texts:

• IM Fell, True Type version (SIL Open Font License, Version 1.1)
• Honoré de Balzac (1830): La Vendetta. Œuvres complètes de H. de Balzac, A. Houssiaux, 1855, 1 (pp. 168-230). https://fr.wikisource.org/wiki/La_Vendetta_%28Balzac%29

ocropy-french-models's Issues

FontForge

http://fontforge.github.io/en-US/

Train also with real examples

E.g. images + text from Œuvres de Frédéric le Grand / UB Trier: http://friedrich.uni-trier.de/de/