tesseract-ocr 0.0.4 → 0.1.0

data/README.md

@@ -12,8 +12,14 @@ because it's still under review for upstream merging.
 
 The gem is called `tesseract-ocr`.
 
-Example
--------
+If you're having any problem requiring tesseract or any dependencies, check the permissions of the installed
+gems.
+
+Examples
+--------
+Following are some examples that show the functionalities provided by tesseract-ocr.
+
+### Basic functionality of tesseract
 
 ```ruby
 require 'tesseract'
@@ -24,26 +30,58 @@ e = Tesseract::Engine.new {|e|
 }
 
 e.text_for('test/first.png').strip # => 'ABC'
-
-e.words_for('test/second.png') # [
-                               #      [ 0] #<Tesseract(93.41653442382812): "|'m">,
-                               #      [ 1] #<Tesseract(91.11811828613281): "12">,
-                               #      [ 2] #<Tesseract(85.71760559082031): "and">,
-                               #      [ 3] #<Tesseract(83.4853515625): "what">,
-                               #      [ 4] #<Tesseract(86.71072387695312): "is">,
-                               #      [ 5] #<Tesseract(83.2227783203125): "this.">,
-                               #      [ 6] #<Tesseract(82.81439208984375): "INSTALL">,
-                               #      [ 7] #<Tesseract(86.46566772460938): "GENTOO">,
-                               #      [ 8] #<Tesseract(93.19613647460938): "OH">,
-                               #      [ 9] #<Tesseract(82.81439208984375): "HAI">,
-                               #      [10] #<Tesseract(85.9158935546875): "1234">
-                               #  ]
 ```
 
 You can pass to `#text_for` either a path, an IO object, a string containing the image or
 an object that responds to `#to_blob` (for example  Magick::Image), keep in mind that
 the format has to be supported by leptonica.
 
+### Accessing advanced features
+
+With advanced features you get access to blocks, paragraphs, lines, words and symbols.
+
+There are lot of way to access those levels, the methods are the following (replace level
+with one of the accessible features, so `each_level` can be `each_block` or `each_paragraph`
+etc.)
+
+The following kind of accessors need a block to be passed and they pass to the block each
+`Element` object. The Element object has various getters to access certain features, I'll
+talk about them later.
+
+The methods are:
+
+* `each_level`
+* `each_level_for`
+* `each_level_at`
+
+The following accessors instead return an `Array` of `Element`s with cached getters, the getters
+are cached beacause the values accessible in the `Element` are linked to the state of the internal
+API, and that state changes if you access something else.
+
+The methods are:
+
+*	`levels`
+*	`levels_for`
+*	`levels_at`
+
+Again, to `*_for` methods you can pass what you can pass to a `#text_for`.
+
+Each `Element` object has the following getters:
+
+* `bounding_box`, this will return the box where the element is confined into
+* `binary_image`, this will return the bichromatic image of the element
+* `image`, this will return the image of the element
+* `baseline`, this will return the line where the text is with a pair of coordinates
+* `orientation`, this will return the orientation of the element
+* `text`, this will return the text of the element
+* `confidence`, this will return the confidence of correctness for the element
+
+`Block` elements also have `type` accessors that specify the type of the block.
+
+`Word` elements also have `font_attributes`, `from_dictionary?` and `numeric?` getters.
+
+`Symbol` elements also have `superscript?`, `subscript?` and `dropcap?` getters.
+
 Using the binary
 ----------------
 You can also use the shipped executable in the following way:

data/bin/tesseract-train.rb

@@ -0,0 +1,63 @@
+#! /usr/bin/env ruby
+require 'optparse'
+require 'tmpdir'
+require 'fileutils'
+require 'shellwords'
+
+options = {}
+
+OptionParser.new do |o|
+	o.on '-b', '--box FILE', 'the box file to use' do |value|
+		options[:box] = File.realpath(value)
+	end
+
+	o.on '-i', '--image FILE', 'the image file to use' do |value|
+		options[:image] = File.realpath(value)
+	end
+
+	o.on '-o', '--output FILE', 'the path where to output the traineddata' do |value|
+		options[:output] = File.expand_path(value)
+	end
+end.parse!
+
+if language = ARGV.shift
+	options[:box]    = File.realpath("#{language}.box")
+	options[:image]  = File.realpath("#{language}.tif")
+	options[:output] = File.expand_path("#{language}.traineddata")
+else
+	language = options[:box][/^(.*?)\./, 1]
+end
+
+Dir.chdir FileUtils.mkpath(File.join(Dir.tmpdir, rand.to_s)).first
+
+language = language.shellescape
+
+%x{
+	cp #{options[:box].shellescape} #{language}.box
+	cp #{options[:image].shellescape} #{language}#{File.extname(options[:image])}
+
+	tesseract #{language}#{File.extname(options[:image])} #{language} nobatch box.train.stderr
+
+	unicharset_extractor #{language}.box
+
+	echo #{language} 0 0 0 0 0 > font_properties
+	mftraining -F font_properties -U unicharset #{language}.tr
+	mftraining -F font_properties -U unicharset -O #{language}.unicharset #{language}.tr
+	cntraining #{language}.tr
+
+	mv Microfeat #{language}.Microfeat
+	mv normproto #{language}.normproto
+	mv pffmtable #{language}.pffmtable
+	mv mfunicharset #{language}.mfunicharset
+	mv inttemp #{language}.inttemp
+
+	combine_tessdata #{language}.
+
+	mv #{language}.traineddata #{options[:output].shellescape}
+}
+
+path = File.realpath(Dir.pwd)
+
+Dir.chdir '/'
+
+FileUtils.rm_rf path

data/examples/nerdz-captcha-breaker/break.rb

@@ -2,21 +2,29 @@
 require 'tesseract'
 require 'RMagick'
 
+# this function is used to get points near the current pixel to
+# cleanup the oblique lines mess, horizontal points seem to output
+# better cleanup
 def near (x, y)
 	[
-		[x - 1, y - 1],
-		[x,     y - 1],
-		[x + 1, y - 1],
+#		[x - 1, y - 1],
+#		[x,     y - 1],
+#		[x + 1, y - 1],
 		[x - 1, y    ],
-		# FIRE IN THE HOLE
 		[x + 1, y    ],
-		[x - 1, y + 1],
-		[x,     y + 1],
-		[x + 1, y + 1]
+#		[x - 1, y + 1],
+#		[x,     y + 1],
+#		[x + 1, y + 1]
 	]
 end
 
-Tesseract::Engine.new.tap {|engine|
+ENV['TESSDATA_PREFIX'] = './'
+
+Tesseract::Engine.new {|engine|
+	engine.language               = :lol
+	engine.page_segmentation_mode = 8
+	engine.whitelist              = [*'a'..'z', *'A'..'Z', *0..9].join
+}.tap {|engine|
 	ARGV.each {|path|
 		image  = Magick::Image.read(path).first
 		pixels = Hash.new { |h, k| h[k] = 0 }
@@ -51,8 +59,8 @@ Tesseract::Engine.new.tap {|engine|
 			image.pixel_color x, y, 'black'
 		}
 
-		File.open('/tmp/lol.png', ?w) { |f| f.write(image.resize(10).to_blob) }
+		image.scale(4).display if ENV['DEBUG']
 
-		puts engine.text_for(image.resize 10).strip
+		puts "#{path}: #{engine.text_for(image.scale(4)).strip}"
 	}
 }

data/examples/nerdz-captcha-breaker/captchas/captchas.txt

@@ -0,0 +1,100 @@
+enKzaV
+CZU6tf
+ZO5mGY
+FNNhZv
+Dwp1Vy
+JYsDAi
+dld510
+yG615j
+WHxAUZ
+k9IhZu
+qWIPSr
+nDSXc5
+9iTYeZ
+s44iQ9
+VPNXWy
+80zxvW
+QA7IYj
+D8Ro4U
+OiEg1U
+pJS7Z8
+6w8eik
+s5igED
+7bJe8p
+VtYdW3
+jNNdcO
+neLPNV
+KONPnl
+Q8aSXJ
+kIwSqv
+8LQExn
+RwcDU2
+LMLg5K
+C0YmdD
+mqAvES
+Ai0Wxi
+bopETp
+L3yP5u
+w4rw3b
+oSEUMU
+bftqDK
+mM7cKE
+rYl6x4
+3hVI8X
+Tm2PPp
+VfmqQ6
+0EZAgC
+QW6gBS
+UTS137
+YXXTqk
+a6LU3K
+SVzguN
+l9G8Y9
+ZP9TDM
+yj7zmS
+sD0Ub9
+XeWY3A
+w8EKVl
+3gO266
+yYN3oJ
+NumjLi
+EEzwCz
+8bUSjW
+GAo6ap
+AXcn6K
+KquWmp
+PM8LYt
+uGS7GO
+mucVzf
+UIZ6cf
+mXRdGq
+lcn8OP
+lKqkw7
+CTHnR1
+ShvmSD
+klCoXI
+8epftU
+nA277p
+uXfavQ
+EhInXB
+KJLUZf
+nAWswu
+savOX1
+bJtoCK
+cTnmnF
+hoYGbS
+d3J04A
+c18WaN
+aO5VVN
+DvP3vf
+7cuFOq
+mPaSmA
+k0rNqQ
+tvG40k
+154xFW
+x9ioCI
+p9rR4J
+DNLPJA
+wGR2O1
+zVr96r
+856sHR