Tesseract font training materials

Tesseract font training materials

1. Create a. Box file.

tesseract [lang].[fontname].exp[num].tif [lang].[fontname].exp[num] -l yournewlanguage batch.nochop makebox

 

2. Start Training

tesseract [lang].[fontname].exp[num].tif [lang].[fontname].exp[num] box.train

Or

tesseract [lang].[fontname].exp[num].tif [lang].[fontname].exp[num] box.train.stderr

Set_unicharset_properties

I don't know what it is.

training/set_unicharset_properties -U input_unicharset -O output_unicharset --script_dir=training/langdata

 

Font_properties

Font property File

<fontname> <italic> <bold> <fixed> <serif> <fraktur>

In <font>, It is a font named by a string; <italic >,< bold >,< fixed>, both <linefeed> and <goth> are simple 0 or 1 signs indicating whether the font is true or not.

Example:

timesitalic 1 0 0 1 0

 

---- In 3.03, there is a default font_properties file that covers 3000 fonts (not necessarily accurate) training/langdata/font_properties.

 

Clustering

Shapeclustering creates the cluster shape of the master shape table and writes it to a file shapetable.

shapeclustering -F font_properties -U unicharset lang.fontname.exp0.tr lang.fontname.exp1.tr ...

---- If you get an error message like "index> = 0 & index <size_used _: Error: assert failed in genericvector. h, line 512 "add the shapetable file to your language data file.

 

 

mftraining -F font_properties -U unicharset -O lang.unicharset lang.fontname.exp0.tr lang.fontname.exp1.tr ...

Your file is the unicharset generated by unicharset_extractor or above, and Lang. unicharset is the output unicharset that will be given to combine_tessdata. Mftraining will output two data files: inttemp (shape prototype) and pffmtable (expected functionality for each character ).

 

Output The normproto Data File

cntraining lang.fontname.exp0.tr lang.fontname.exp1.tr ...

 

 

Data Dictionary (optional)

 

Name	Type	Description
Word-dawg	Dawg	A Dawg made from dictionary words from the language.
Freq-dawg	Dawg	A Dawg made from the most frequent words which wowould have gone into word-dawg.
Punc-dawg	Dawg	A Dawg made from punctuation patterns found around words. The "word" part is replaced by a single space.
Number-dawg	Dawg	A Dawg made from tokens which originally contained digits. Each digit is replaced by a space character.
Fixed-length-Dawgs	Dawg	Several dawgs of different fixed lengths -- useful for different ages like Chinese.
Bigram-dawg	Dawg	A Dawg of word bigrams where the words are separated by a space and each digit is replaced by ?.
Unambig-dawg	Dawg	Todo: describe.
User-Words	Text	A list of extra words to add to the dictionary. Usually left empty to be added by users if they require it; see tesseract (1 ).

wordlist2dawg frequent_words_list lang.freq-dawg lang.unicharset
wordlist2dawg words_list lang.word-dawg lang.unicharset

 

References:

Wiki

Https://code.google.com/p/tesseract-ocr/wiki/FAQ

 

Introduction

Https://code.google.com/p/tesseract-ocr/wiki/TrainingTesseract3#font_properties_ (new_in_3.01)

 

Wordlist2dawg (1) manual page

Http://tesseract-ocr.googlecode.com/svn-history/trunk/doc/wordlist2dawg.1.html

 

Combine_tessdata (1) manual page

Http://tesseract-ocr.googlecode.com/svn-history/r800/trunk/doc/combine_tessdata.1.html