Skip to content

Latest commit

 

History

History
218 lines (192 loc) · 8.25 KB

README.md

File metadata and controls

218 lines (192 loc) · 8.25 KB
Raw

TL;DR If you just want to use this executable and not go to the trouble of having to build it, please skip ahead to this section below.

Installing Tesseract 4.0.0 on Centos7

The goal of this repo is to show how to use a CentOS7 system (with root access), to create a static compiled binary which can be copied over to, and used on, a CentOS7 system (without root access).

Question: Why would we want to do this?

Answer: In some cases you might want to use tesseract on a machine via a cloud provider. For security reasons, specific machines on a specific cloud provider's infrastructure will not allow root access to the remove guest.

Solution: What we are about to do is log into a CentOS7 machine with root access (this can be any machine i.e. a VM on your local machine etc.). We install all of the dependencies and create sufficuent executables on the machine with root access.

Then, we copy the tesseract executable over to the machine without root access. We also make sure that the executable is in the path. We also make sure that we have the correct trained language data. Finally we fetch an image and perform the OCR.

The CentOS7 machine WITH root access

For this task you can use your own CentOS machine etc. This will have to be a machine on which you have full priveledges.

Install system dependencies

sudo yum install -y git
sudo yum install -y libtool
sudo yum install clang
sudo yum install -y gcc-c++.x86_64

Install leptonica dependencies

sudo yum install -y zlib
sudo yum install -y zlib-devel
sudo yum install -y libjpeg
sudo yum install -y libjpeg-devel
sudo yum install -y libwebp
sudo yum install -y libwebp-devel
sudo yum install -y libtiff
sudo yum install -y libtiff-devel
sudo yum install -y libpng
sudo yum install -y libpng-devel

Move executables to /usr/local/lib

cd /usr/local/lib
sudo cp /usr/lib64/libjpeg.so.62 .
sudo cp /usr/lib64/libwebp.so.4 .
sudo cp /usr/lib64/libtiff.so.5 .
sudo cp /usr/lib64/libpng15.so.15 .

Install leptonica from source

cd /home/azureuser
git clone https://github.com/DanBloomberg/leptonica.git --depth 1
cd /home/azureuser/leptonica
./autogen.sh
./configure --prefix=/usr/local --disable-shared --enable-static --with-zlib --with-jpeg --with-libwebp  --with-libtiff --with-libpng
make
sudo make install
sudo ldconfig

Install tesseract from source

cd /home/azureuser
wget https://github.com/tesseract-ocr/tesseract/archive/4.0.0.tar.gz -O tesseract-4.0.0.tar.gz
tar zxvf tesseract-4.0.0.tar.gz
cd tesseract-4.0.0
export PKG_CONFIG_PATH=/usr/local/lib/pkgconfig
./autogen.sh
./configure --prefix=/usr/local --disable-shared --enable-static --with-extra-libraries=/usr/local/lib/ --with-extra-includes=/usr/local/lib/
make
sudo make install
sudo ldconfig

The CentOS7 machine WITHOUT root access

Getting the executable

Option 1 - Use your own executable (using the procedure above)

If you followed the procedure above, please copy the tesseract binary (from the machine with root access) to the Centos7 machine without root access.

cd ~
mkdir -p /home/azureuser/tess
cd /home/azureuser/tess
scp -i ~/.ssh/key.pem -rp azureuser@xxx.xx.xxx.xxx:/usr/local/bin/tesseract .

Then ensure that the tesseract binary is in the system path on the machine without root access, by adding the export statement to the ~/.bash_profile file

export PATH="$PATH:/home/azureuser/tess"

Then action those changes

source ~/.bash_profile

Option 2 - Use our pre-made executable (that we prepared earlier)

If you just want to use a pre-made executable, you can use wget to fetch it like this. First log into the CentOS7 machine without root access. Then fetch the executable like this.

cd ~
mkdir -p /home/azureuser/tess
cd /home/azureuser/tess
wget https://github.com/second-state/OCR-tesseract-on-Centos7/raw/main/tesseract.tar.gz
tar -zxvf tesseract.tar.gz

Then ensure that the tesseract binary is in the system path on the machine without root access, by adding the export statement to the ~/.bash_profile file

export PATH="$PATH:/home/azureuser/tess"

Then action those changes

source ~/.bash_profile

Finalizing dependencies

Option 1 - use your own files

Create a new deps directory on the CentOS7 without root access

mkdir -p /home/azureuser/tess/deps

Take the following files from the CentOS7 with root access and put them in the deps directory that you just created.

/usr/lib64/libjbig.so.2.0
/usr/local/lib/libwebp.so.4
/usr/local/lib/libtiff.so.5
/usr/local/lib/libjpeg.so.62

Then the following line to the ~/.bash_profile file on the CentOS7 without root access

export LD_LIBRARY_PATH="/home/azureuser/tess/deps/"

Then action those changes

source ~/.bash_profile

Option 2 - use our files (that we prepared earlier)

cd /home/azureuser/tess
wget https://github.com/second-state/OCR-tesseract-on-Centos7/raw/main/deps.tar.gz
tar -zxvf deps.tar.gz

Then the following line to the ~/.bash_profile file on the CentOS7 without root access

export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/home/azureuser/tess/deps/"

Then action those changes

source ~/.bash_profile

Load tesseract languages

Create a new location to store the trained language data and then export that location as the TESSDATA_PREFIX

mkdir -p /home/azureuser/tess/traineddata

Then add this path to your ~/.bash_profile file.

export TESSDATA_PREFIX=/home/azureuser/tess/traineddata

Then action those changes

source ~/.bash_profile

Copy any of the .traineddata files available here that you need to the above TESSDATA_PREFIX location.

cd $TESSDATA_PREFIX
wget https://github.com/tesseract-ocr/tessdata/raw/master/fra.traineddata
wget https://github.com/tesseract-ocr/tessdata/raw/master/eng.traineddata

Try tesseract - using the system's environment variables

Grab an image such as [this french lunch menu example]( wget https://second-state.github.io/wasm-learning/faas/ocr/html/a_french_lunch_menu.png)

cd /home/azureuser/tess
wget https://second-state.github.io/wasm-learning/faas/ocr/html/a_french_lunch_menu.png

Then run the tesseract command and pass in the image as the first parameter

cd /home/azureuser/tess
./tesseract a_french_lunch_menu.png stdout --dpi 70 -l fra

Try tesseract - manually passing environment vars as part of the command

In cases where you are calling the tesseract as a system command (via a third party environment) you could also not worry about setting environment variables in files like bash_profile etc. Instead you can just make the PATH, TESSDATA_PREFIX, and LD_LIBRARY_PATH part of the actual command.

PATH="$PATH:/home/azureuser/tess" TESSDATA_PREFIX="/home/azureuser/tess/traineddata" LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/home/azureuser/tess/deps/" tesseract a_french_lunch_menu.png stdout --dpi 70 -l fra

This is actually safer than having an LD_LIBRARY_PATH set globally.

Potential errors

If you get an error relating to a dependency i.e.

./tesseract: error while loading shared libraries: xxxxxxx.so.x: cannot open shared object file: No such file or directory

Please run ldd tesseract as shown below to see which dependencies are => not_found and then place an issue so I can add those to the deps.tar.gz

[azureuser@centos7-ocr tess]$ ldd tesseract
	linux-vdso.so.1 =>  (0x00007fff2c5bd000)
	libpng15.so.15 => /lib64/libpng15.so.15 (0x00007f55715d6000)
	libjpeg.so.62 (0x00007f5571381000)
	libtiff.so.5 (0x00007f557110d000)
	libwebp.so.4 (0x00007f5570ebd000)
	libz.so.1 => /lib64/libz.so.1 (0x00007f5570ca7000)
	librt.so.1 => /lib64/librt.so.1 (0x00007f5570a9f000)
	libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f5570883000)
	libstdc++.so.6 => /lib64/libstdc++.so.6 (0x00007f557057c000)
	libm.so.6 => /lib64/libm.so.6 (0x00007f557027a000)
	libgomp.so.1 => /lib64/libgomp.so.1 (0x00007f5570054000)
	libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007f556fe3e000)
	libc.so.6 => /lib64/libc.so.6 (0x00007f556fa70000)
	libjbig.so.2.0 => /home/azureuser/tess/deps/libjbig.so.2.0 (0x00007f556f864000)
	/lib64/ld-linux-x86-64.so.2 (0x00007f5571801000)