A toy command line utility for OCRing and cleaning OCR output.

tesse

#!/usr/bin/env ruby

# tesse: commandline tool for looking at tesseract OCR and cleaning the output

# Besides the following gem requirements it requires the following Linux programs:
# eog: for viewing the images
# wmctrl: for resizing and positioning the image viewing window

require 'tesseract'
require 'ffi/aspell'
require 'tempfile'
require 'formatador'

file = File.expand_path(ARGV[0])
system('clear')

engine = Tesseract::Engine.new {|e|
  e.language  = :eng
  e.blacklist = '|'
  e.whitelist = [*'a'..'z', *'A'..'Z', *0..9, " ."].join
}

Speller = FFI::Aspell::Speller.new('en_US')

def spelling_percentage(text)
  words = text.split(' ')
  total = words.length
  correct = 0.0
  words.each do |word|
    if Speller.correct?(word)
      correct += 1
    end
  end
  correct / total * 100.0
end

cleaned_ocr = []
engine.each_line_for(file) do |line|
  text = line.text
  puts text
  Formatador.display "[green]\n-----------------------------------------------\n[/]"
  table = {}
  table[:confidence] = line.confidence
  table[:spelling] = spelling_percentage(text)
  Formatador.display_table([table])

  # write the line image to a file
  file = Tempfile.new('tesse_image')
  # puts file.path
  file.write line.image.to_blob

  # show the image in eog and position the windows nicely
  eog_pid = fork do
    exec %Q|eog #{file.path} 2>/dev/null|
  end
  system "sleep 1; wmctrl -a tesse; wmctrl -r tesse_image -e 0,0,700,1500,300;"

  print "Continue? "
  value = STDIN.gets.chomp

  case value
  when 'e'
    puts "Type out a better transcript of this line:"
    text = STDIN.gets.chomp
  when 'd'
    text = nil
  end

  cleaned_ocr << text

  # clean up by killing the image viewer process 
  Process.kill 'ABRT', eog_pid
  `killall eog`
  system('clear')
  file.close
  file.unlink  
end

puts 'Done! Following is the cleaned output'
Formatador.display "[green]\n-----------------------------------------------\n[/]"
cleaned_ocr.compact!
puts cleaned_ocr