OCR with Tesseract¶

In order to run Tesseract, the below Installation steps have to be performed.

Download & Install the Tesseract Language Data files¶

Download and Install the tesseract language data files for Version 3.X on each of the worker nodes of the cluster:

https://github.com/tesseract-ocr/tessdata/releases
wget https://github.com/tesseract-ocr/tessdata/archive/3.04.00.tar.gz

Install them in the same directory on each of the worker nodes:
```
git clone https://github.com/tesseract-ocr/tessdata.git
```

Include TESSDATA_PREFIX in spark configs when submitting the job¶

Include the following in spark submit configs when running workflows containing the OCR node:
```
--conf spark.executorEnv.TESSDATA_PREFIX=/home/ec2-user/tessdata
```
Where the tesseract language data files are in /home/ec2-user/tessdata directory on each of the worker nodes

Error if TESSDATA_PREFIX is not set correctly¶

If TESSDATA_PREFIX is not set, the spark program would run into the error below:

Error opening data file /Users/saudet/projects/bytedeco/javacpp-presets/tesseract/cppbuild/macosx-x86_64/share/tessdata/eng.traineddata
Please make sure the TESSDATA_PREFIX environment variable is set to the parent directory of your "tessdata" directory.
Failed loading language 'eng'
Tesseract couldn't load any languages!

The above error would be in the Job logs. If yarn is being used it would be in the yarn logs:

yarn logs -applicationId job_application_id

When the job is being executed, Fire displays the job_application_id in the browser.