Tesseract on 100s, 1000s of Files

The real benefit to using Tesseract over other OCR tools is the speed and ease of processing batches of files.

Basic input:

From within the directory of files (in this case, .jpg files). For each .jpg:

• Echo the name of the file (to keep track of what is being processed)
• Instruct tesseract to run on a page, generate an image output, and apply any parameters.
• complete the loop with “done”

Input:

for page in *.jpg
do 
    echo "$page"
    tesseract "$page" "$page-output" -l eng
done

Did it work? First check if the files have been created by listing everything in the folder:

Input:

ls

You should have a .txt file for every image file in the directory. The next step is to figure out if tesseract actually ran and wrote text to each file. Run head to check.

Input:

head *.txt

Output example:

==> 00000012.jpg-output.txt <==
2 A CHRISTMAS CAROL.

hands shall not disturb it, or the Country ’s done for.
You will therefore permit me to repeat, emphati-
cally, that Marley was as dead as a door-nail.

Scrooge knew he was dead? Of course he did.
How could it be otherwise? Scrooge and he were
partners for I don’t know how many years. Scrooge
was his sole executor, his sole administrator, his

==> 00000013.jpg-output.txt <==
MARLEY'S GHOST. 3

would be in any other middle-aged gentleman rashly
turning out after dark in a breezy spot—say Saint
Paul’s Churchyard for instance—literally to astonish
his son’s weak mind.

Scrooge never painted out Old Marley’s name.
There it stood, years afterwards, above the ware-
house door: Scrooge and Marley. The firm was

If you are running OCR on batches of random images, separate text files are useful. However, if you have a single book, you probably want the full text bundled together.

We can combine cat and an output redirect to combine the files.

Input:

cat *.txt > full-book-title.txt

If it has run correctly, there will not be any message. To check if the file has been written correctly, we can run cat, less, head, or tail.