pytesstrain 0.1.1

Collection of utilities for Tesseract OCR training

Python utilities for Tesseract OCR training

This module is a collection of different training utilities for Tesseract OCR. These utilities are also implemented as console scripts, hence they can be run from command line.

Utilities

All utilities list their command line switches when run with the switch --help.

• rewrap just rewraps text lines by specified maximal line length
• create_dictdata creates all word- and n-gram-lists from a text file, which are translated to DAWGs and added to the traineddata file then
• language_metrics creates random texts from supplied wordlist and tests for recognition error rates
• collect_ambiguities extracts error-correction pairs from reference-hypothesis pairs and stores them in a JSON file
• json2unicharambigs stores specified error-correction pairs from JSON file in a unicharambigs file

Requirements

This module requires the following modules to work:

• pytesseract (Running Tesseract OCR)
• editdistance (Calculation of error rates)

Packages

The module is split in several packages. The package pytesstrain.train contains the workhorse function run_text(). The package pytesstrain.cli contains the utilities you might run at the command line. The package pytesstrain.ambigs contains function around unicharambigs file. The package pytesstrain.text2image contains the interface to the text2image command from the Tesseract OCR; the interface relies on pytesseract module and is modelled after it as well. The package pytesstrain.metrics contains error rate calculations, as well the interface class Metrics. The package pytesstrain.utils contains auxiliary functions.