Shell script that takes a pdf of scanned text, and uses the Tesseract OCR library to produce a text file

pdf2txt.sh

#!/bin/bash
#
# Takes pdf of scanned text, and uses the Tesseract OCR library to produce 
# a text version.

set -e
[ -z "$1" ] && echo "USAGE: pdf2txt INPUT_PATH [OUTPUT_PATH]" && exit 1

input="$1"

tmpdir=$(mktemp -d /tmp/ocr.XXXX)

pdftk "$input" dump_data | grep NumberOfPages

echo 'splitting up pdf...'
pdftk "$input" burst output $tmpdir/%04d.pdf
rm $tmpdir/doc_data.txt

for f in $tmpdir/*.pdf
do
  echo "Converting $f"
  convert -density 300 "$f" -quality 90 -limit memory 3GB -limit disk 10GB ${f%.*}.png
done
echo 'Done converting'

#convert -density 300 "$input" -quality 90 -limit memory 3GB -limit disk 10GB $tmpdir/%04d.png

for f in $tmpdir/*.png
do
  echo "Reading $f"
  tesseract $f ${f%.*}.txt > /dev/null 2>&1
done

if [ -z "$2" ]
  then
    output=${input%.*}.txt
  else
    output=$2
fi

echo "" > $output
counter=1
for f in $tmpdir/*.txt
do
  echo -e "\n## Page $counter\n" >> $output
  cat $f >> $output
  counter=$((counter+1))
done

rm $tmpdir/*
rmdir $tmpdir