Building tesserocr on MS Windows 64bit

If you search for efficient solution of using tesseract OCR in python you will need to use tessocer. But there are no recent version of project for current version of python on windows. So you have build it by yourself.

Requirements

• cmake
• Visual Studio Community
• unzip (part of git for windows) or use windows build-in solutions
• git for windows (optional)
• curl for downloading files from internet (optional)

Installing dependencies

Project structure and initialisation of environment

Project will be created on drive "F:" You can adjust this path to your needs.

mkdir F:\win64
set INSTALL_DIR=F:\win64
set PATH=%PATH%;%INSTALL_DIR%\bin

mkdir %INSTALL_DIR%\share\tesseract\tessdata

Source tree we will put its own place:

mkdir F:\Project

Initialize VS environment:

call "c:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Auxiliary\Build\vcvars64.bat" x64

set INCLUDE=%INCLUDE%;%INSTALL_DIR%\include
set LIBPATH=%LIBPATH%;%INSTALL_DIR%\lib
set TESSDATA_PREFIX=%INSTALL_DIR%\share\tesseract\tessdata

tessdata

This is place where the language data files are located together with common configurations. We will use tessdata repository as in contains models for legacy and LSTM engine. You can also download data from tessdata_best or tessdata_fast repository, but they have only LSTM model.

Legacy engine is better if you need to find exact location of individual characters, but has not so good accuracy as LSTM engine. Therefore some users suggest combination of both engines.

git clone --depth 1 https://github.com/tesseract-ocr/tessconfigs %TESSDATA_PREFIX%
curl -L https://github.com/tesseract-ocr/tessdata/raw/master/eng.traineddata ^
    --output %TESSDATA_PREFIX%\eng.traineddata
curl -L https://github.com/tesseract-ocr/tessdata/raw/master/osd.traineddata ^
    --output %TESSDATA_PREFIX%\osd.traineddata

Building tesseract and leptonica with dependencies

We will install only zlib and libpng because we will use Python  Pillow for image opening. But you can install also other dependencies (gif, jpeg, tiff, webp) before building leptonica

zlib build and installation

F:
cd F:\Project
curl https://zlib.net/zlib1211.zip
"c:\Program Files\Git\usr\bin\unzip.exe" zlib1211.zip
cd zlib-1.2.11
cmake -Bbuild -DCMAKE_BUILD_TYPE=Release ^
    -DCMAKE_PREFIX_PATH=%INSTALL_DIR% ^
    -DCMAKE_INSTALL_PREFIX=%INSTALL_DIR%
cmake --build build --config Release --target install
cd ..

libpng build and installation

curl https://vorboss.dl.sourceforge.net/project/libpng/libpng16/1.6.37/lpng1637.zip
"c:\Program Files\Git\usr\bin\unzip.exe" lpng1637.zip
cd lpng1637
cmake -Bbuild -DCMAKE_BUILD_TYPE=Release ^
    -DCMAKE_PREFIX_PATH=%INSTALL_DIR% -DCMAKE_INSTALL_PREFIX=%INSTALL_DIR%
cmake --build build --config Release --target install
cd ..

leptonica build and installation

git clone --depth 1 https://github.com/DanBloomberg/leptonica.git
cd leptonica
cmake -Bbuild -DCMAKE_BUILD_TYPE=Release ^
   -DCMAKE_INSTALL_PREFIX=%INSTALL_DIR% -DCMAKE_PREFIX_PATH=%INSTALL_DIR% ^
   -DBUILD_PROG=OFF -DSW_BUILD=OFF -DBUILD_SHARED_LIBS=ON
cmake --build build  --config Release --target install
cd ..

tesseract build and installation

We will use the latest stable version as master (a.k.a 5.0 version) API is not stable and changing at the moment.

git clone -b 4.1.1 --depth 1 https://github.com/tesseract-ocr/tesseract.git
cd tesseract
cmake -Bbuild -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=%INSTALL_DIR% ^
    -DCMAKE_PREFIX_PATH=%INSTALL_DIR% ^
    -DLeptonica_DIR=%INSTALL_DIR%\lib\cmake  ^
    -DBUILD_TRAINING_TOOLS=OFF -DSW_BUILD=OFF ^
    -DOPENMP_BUILD=OFF -DBUILD_SHARED_LIBS=ON
cmake --build build --config Release --target install
cd ..

tesserocr build

git clone --depth 1 https://github.com/sirfz/tesserocr.git
cd tesserocr
python setup.py build
python setup.py bdist_wheel
pip uninstall -y tesserocr
pip install dist\tesserocr-2.5.2b0-cp38-cp38-win_amd64.whl
copy F:\win64\bin\*.dll "C:\Program Files\Python38\Lib\site-packages\"
(adjust based on your Python installation)
tests
cd F:\Project\tesserocr
python
>>> import tesserocr
>>> tesserocr.PyTessBaseAPI.Version()
'4.1.1'
>>> tesserocr.get_languages()
('F:\\Project\\tessdata/', ['eng', 'osd'])
>>> from PIL import Image
>>> image = Image.open(r'F:\Project\tesserocr\tests\eurotext.png')
>>> with tesserocr.PyTessBaseAPI() as api:
...     api.SetImage(image)
...     print(api.GetUTF8Text())
...
The (quick) [brown] {fox} jumps!
Over the $43,456.78 <lazy> #90 dog
& duck/goose, as 12.5% of E-mail
from aspammer@website.com is spam.
Der ,schnelle” braune Fuchs springt
iiber den faulen Hund. Le renard brun
«rapide» saute par-dessus le chien
paresseux. La volpe marrone rapida
salta sopra il cane pigro. El zorro
marron ripido salta sobre el perro
perezoso. A raposa marrom ripida
salta sobre o cdo preguigoso.
>>>