Trains a multi-layer perceptron (MLP) neural network to perform optical character recognition (OCR).

The training set is automatically generated using a heavily modified version of the captcha-generator node-captcha. Support for the MNIST handwritten digit database has been added recently (see performance section).

The network takes a one-dimensional binary array (default 20 * 20 = 400-bit) as input and outputs an 8-bit array, which can then be converted into a character code. Initial performance measurements show promising success rates.

After training, the network is saved as a standalone module to ./ocr.js, which can then be used in your project with

var ocr = require('./ocr.js');
var output = ocr
  .activate(input)
  .map(function(bit) {
    return bit > 0.5 ? 1 : 0;
  });
  
// output is now binary array, converting to character is easy
var character = String.fromCharCode(parseInt(output.join(''), 2));

// do stuff ...

• Neurons
  • 400 input
  • 40 hidden
  • 4 output
• Learning rate
  • 0.1
• Training set
  • 59999 digits
• Testing set
  • 9999 digits
• Measured success rate
  • 82.08820882088209%

• Fonts
  • sans-serif
  • serif
• Neurons
  • 400 input
  • 40 hidden
  • 8 output
• Learning rate
  • 0.1
• Training set
  • 52000 characters
  • Sample
    • 
• Testing set
  • 13000 characters
• Measured success rate
  • 96.32307692307693%

• Fonts
  • sans-serif
  • serif
• Neurons
  • 400 input
  • 40 hidden
  • 8 output
• Learning rate
  • 0.1
• Training set
  • Size
    • 20000 digits
  • Sample
    • 
• Testing set
  • 5000 digits
• Measured success rate
  • 99.22%

Tweak the network for your needs by editing the config.json file located in the main folder. Pasted below is the default config file.

{
  "mnist": false,
  "text": "abcdefghijklmnopqrstuvwxyz",
  "fonts": [
    "sans-serif",
    "serif"
  ],
  "training_set": 2000,
  "testing_set": 500,
  "image_size": 20,
  "threshold": 400,
  "network": {
    "hidden": 40,
    "learning_rate": 0.1
  }
}

• mnist
  • If set to true, the MNIST handwritten digit dataset will be used for training and testing the network. This setting will overwrite configured set sizes and will ignore the image_size, threshold, fonts and text settings.
• text
  • A string containing the glyphs with which to train/test the network.
• fonts
  • An array of fonts to be used when generating images.
• training_set
  • Number of images to be generated and used as the network training set.
• testing_set
  • Same as above, but these images are used for testing the network.
• image_size
  • The size of the square chunk (in pixels) containing a glyph. The resulting network input size is image_size^2.
• threshold
  • When analyzing the pixels of a glyph, the algorithm reduces each pixel (r, g, b) to (r + g + b) and everything below threshold is marked as 1 in the resulting binary array used as network input.
• network
  • hidden
    • The size (number of neurons) of the hidden layer of the network.
  • learning_rate
    • The learning rate of the network.

Clone this repository. The script is using canvas, so you'll need to install the Cairo rendering engine. On OSX, this can be done with the following one-liner (copied from canvas README):

$ wget https://raw.githubusercontent.com/LearnBoost/node-canvas/master/install -O - | sh

Then install npm dependencies and test it:

$ npm install
$ node main.js

Here is an example run of the script (with mnist set to true in config.json):

$ node main.js
reading config file ...
... done

parsing MNIST data ...
digit 6 from training set
00000000000000000000
00000001111000000000
00000011110000000000
00000011100000000000
00000111000000000000
00000111000000000000
00001110000000000000
00001110000000000000
00001110000000000000
00001100000001110000
00011100000111110000
00011100001111111000
00011100001110111000
00011100011100111000
00001110111000111000
00001111111001110000
00000111111111110000
00000011111111000000
00000001111110000000
00000000000000000000

digit 8 from testing set
00000000000000000000
00000000111110000000
00000001111111110000
00000011110011111100
00000011100000111100
00000011100000011100
00000011100000111100
00000001100001111000
00000001111111110000
00000011111111100000
00001111111110000000
00001111111100000000
00011100111100000000
00011000111100000000
00111000011000000000
00111000011000000000
00111000011000000000
00011100011000000000
00011100111000000000
00001111110000000000

... done

neural network specs:
  layers:
    input: 400 neurons.
    hidden: 40 neurons.
    output: 4 neurons.
  learning rate: 0.1
  training set: 59999 characters.
  testing set: 999 characters.

learning ...
progress: 10%
progress: 20%
progress: 30%
progress: 40%
progress: 50%
progress: 60%
progress: 70%
progress: 80%
progress: 90%
... done

network saved to ./network.js

testing on 999 samples ...
progress: 10%
progress: 20%
progress: 30%
progress: 40%
progress: 50%
progress: 60%
progress: 70%
progress: 80%
progress: 90%
... done

success rate: 81.38138138138137 %