Recursively perform OCR on images, output searchable PDFs

img2pdf.sh

#!/bin/bash

# Current directory contains a folder named 'img', 
# which contains images nested in subfolders, e.g.
#
# img
#   folder1
#     img1.jpg
#     img2.jpg
#   folder2
#     img3.jpg
#
# For each image in img, create a searchable PDF in pdf.
#
# pdf
#   folder1
#     img1.jpg.pdf
#     img2.jpg.pdf
#   folder2
#     img3.jpg.pdf
#
# Requires tesseract.

find img -type f | while read i; 
do 
  o="${i/img/pdf}"; 
  d=`dirname "${o}"`; 
  mkdir -p "${d}"; 
  n="${o}.pdf"; 
  echo "${n}"; 
  tesseract "${i}" stdout pdf > "${n}"; 
done