Indic Tesseract

Repository containing tessdata, source training data and other materials and hacks for improving accuracy of Indic Languages recognition (initially Malayalam) in Tesseract OCR Engine.

Tesseract

Tesseract was originally developed at Hewlett-Packard Laboratories Bristol and at Hewlett-Packard Co, Greeley Colorado between 1985 and 1994, with some more changes made in 1996 to port to Windows, and some C++izing in 1998.

In 2005 Tesseract was open sourced by HP. Since 2006 it is developed by Google.

Installation

You may need to grab some dependecies.

For Ubuntu or Debian,

sudo apt-get install libleptonica-dev libicu-dev libcairo-dev libpango1.0-dev automake libtool libtiff5-dev autoconf pkg-config libpng-dev libjpeg-dev zlib1g-dev

Now download the latest stable release of Tesseract source from here.

Extract the zip and cd into that folder. Now run,

./autogen.sh
./configure
make
sudo make install
make training
sudo make training-install
sudo ldconfig

Take control over your tessdata directry.

sudo chown `whoami` -R /usr/local/share/tessdata

Running

Tesseract needs language data for character regognition.

Clone this repository.

git clone https://github.com/tvsijin/indic-tesseract.git

Copy the contents of tessdata to your local tessdata folder.

cd indic-tesseract
cp tessdata/* /usr/local/share/tessdata/

Now you can run tesseract by,

 tesseract imagename outputbase [-l lang] [-psm pagesegmode] [configfile...]

So basic usage to do OCR on an image called 'myscan.tiff' which contains Malayalam characters, and save the result to 'out.txt' would be,

tesseract myscan.tiff out -l mal

Training

@TODO

Testing

@TODO

Contribution

@TODO

License

Licensed under the Apache License, Version 2.0.