// Cloned by Samaksh Chandra on 5 Dec 2021 from World "Character recognition neural network" by "Coding Train" project
// Please leave this clone trail here.
// Port of Character recognition neural network from here:
// https://github.com/CodingTrain/Toy-Neural-Network-JS/tree/master/examples/mnist
// with many modifications
// --- defined by MNIST - do not change these ---------------------------------------
const PIXELS = 28; // images in data set are tiny
const PIXELSSQUARED = PIXELS * PIXELS;
// number of training and test exemplars in the data set:
const NOTRAIN = 60000;
const NOTEST = 10000;
//--- can modify all these --------------------------------------------------
// no of nodes in network
const noinput = PIXELSSQUARED;
const nohidden = 64;
const nooutput = 10;
const learningrate = 0.1; // default 0.1
// should we train every timestep or not
let do_training = true;
// how many to train and test per timestep
const TRAINPERSTEP = 30;
const TESTPERSTEP = 5;
// multiply it by this to magnify for display
const ZOOMFACTOR = 7;
const ZOOMPIXELS = ZOOMFACTOR * PIXELS;
// 3 rows of
// large image + 50 gap + small image
// 50 gap between rows
const canvaswidth = ( PIXELS + ZOOMPIXELS ) + 50;
const canvasheight = ( ZOOMPIXELS * 3 ) + 100;
const DOODLE_THICK = 18; // thickness of doodle lines
const DOODLE_BLUR = 3; // blur factor applied to doodles
let mnist;
// all data is loaded into this
// mnist.train_images
// mnist.train_labels
// mnist.test_images
// mnist.test_labels
let nn;
let trainrun = 1;
let train_index = 0;
let testrun = 1;
let test_index = 0;
let total_tests = 0;
let total_correct = 0;
// images in LHS:
let doodle, demo;
let doodle_exists = false;
let demo_exists = false;
let mousedrag = false; // are we in the middle of a mouse drag drawing?
// save inputs to global var to inspect
// type these names in console
var train_inputs, test_inputs, demo_inputs, doodle_inputs;
// Matrix.randomize() is changed to point to this. Must be defined by user of Matrix.
function randomWeight()
{
return ( AB.randomFloatAtoB ( -0.5, 0.5 ) );
// Coding Train default is -1 to 1
}
// make run header bigger
AB.headerCSS ( { "max-height": "95vh" } );
//--- start of AB.msgs structure: ---------------------------------------------------------
// We output a serious of AB.msgs to put data at various places in the run header
var thehtml;
// 1 Doodle header
thehtml = "<hr> <h1> 1. Doodle </h1> Top row: Doodle (left) and shrunk (right). <br> " +
" Draw your doodle in top LHS. <button onclick='wipeDoodle();' class='normbutton' >Clear doodle</button> <br> ";
AB.msg ( thehtml, 1 );
// 2 Doodle variable data (guess)
// 3 Training header
thehtml = "<hr> <h1> 2. Training </h1> Middle row: Training image magnified (left) and original (right). <br> " +
" <button onclick='do_training = false;' class='normbutton' >Stop training</button> <br> ";
AB.msg ( thehtml, 3 );
// 4 variable training data
// 5 Testing header
thehtml = "<h3> Hidden tests </h3> " ;
AB.msg ( thehtml, 5 );
// 6 variable testing data
// 7 Demo header
thehtml = "<hr> <h1> 3. Demo </h1> Bottom row: Test image magnified (left) and original (right). <br>" +
" The network is <i>not</i> trained on any of these images. <br> " +
" <button onclick='makeDemo();' class='normbutton' >Demo test image</button> <br> ";
AB.msg ( thehtml, 7 );
// 8 Demo variable data (random demo ID)
// 9 Demo variable data (changing guess)
const greenspan = "<span style='font-weight:bold; font-size:x-large; color:darkgreen'> " ;
//--- end of AB.msgs structure: ---------------------------------------------------------
function setup()
{
createCanvas ( canvaswidth, canvasheight );
doodle = createGraphics ( ZOOMPIXELS, ZOOMPIXELS ); // doodle on larger canvas
doodle.pixelDensity(1);
// JS load other JS
// maybe have a loading screen while loading the JS and the data set
AB.loadingScreen();
$.getScript ( "/uploads/codingtrain/matrix.js", function()
{
$.getScript ( "/uploads/codingtrain/nn.js", function()
{
$.getScript ( "/uploads/codingtrain/mnist.js", function()
{
console.log ("All JS loaded");
nn = new NeuralNetwork( noinput, nohidden, nooutput );
nn.setLearningRate ( learningrate );
loadData();
});
});
});
}
// load data set from local file (on this server)
function loadData()
{
loadMNIST ( function(data)
{
mnist = data;
console.log ("All data loaded into mnist object:")
console.log(mnist);
AB.removeLoading(); // if no loading screen exists, this does nothing
});
}
function getImage ( img ) // make a P5 image object from a raw data array
{
let theimage = createImage (PIXELS, PIXELS); // make blank image, then populate it
theimage.loadPixels();
for (let i = 0; i < PIXELSSQUARED ; i++)
{
let bright = img[i];
let index = i * 4;
theimage.pixels[index + 0] = bright;
theimage.pixels[index + 1] = bright;
theimage.pixels[index + 2] = bright;
theimage.pixels[index + 3] = 255;
}
theimage.updatePixels();
return theimage;
}
function getInputs ( img ) // convert img array into normalised input array
{
let inputs = [];
for (let i = 0; i < PIXELSSQUARED ; i++)
{
let bright = img[i];
inputs[i] = bright / 255; // normalise to 0 to 1
}
return ( inputs );
}
function trainit (show) // train the network with a single exemplar, from global var "train_index", show visual on or off
{
let img = mnist.train_images[train_index];
let label = mnist.train_labels[train_index];
// optional - show visual of the image
if (show)
{
var theimage = getImage ( img ); // get image from data array
image ( theimage, 0, ZOOMPIXELS+50, ZOOMPIXELS, ZOOMPIXELS ); // magnified
image ( theimage, ZOOMPIXELS+50, ZOOMPIXELS+50, PIXELS, PIXELS ); // original
}
// set up the inputs
let inputs = getInputs ( img ); // get inputs from data array
// set up the outputs
let targets = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0];
targets[label] = 1; // change one output location to 1, the rest stay at 0
// console.log(train_index);
// console.log(inputs);
// console.log(targets);
train_inputs = inputs; // can inspect in console
nn.train ( inputs, targets );
thehtml = " trainrun: " + trainrun + "<br> no: " + train_index ;
AB.msg ( thehtml, 4 );
train_index++;
if ( train_index == NOTRAIN )
{
train_index = 0;
console.log( "finished trainrun: " + trainrun );
trainrun++;
}
}
function testit() // test the network with a single exemplar, from global var "test_index"
{
let img = mnist.test_images[test_index];
let label = mnist.test_labels[test_index];
// set up the inputs
let inputs = getInputs ( img );
test_inputs = inputs; // can inspect in console
let prediction = nn.predict(inputs); // array of outputs
let guess = findMax(prediction); // the top output
total_tests++;
if (guess == label) total_correct++;
let percent = (total_correct / total_tests) * 100 ;
thehtml = " testrun: " + testrun + "<br> no: " + total_tests + " <br> " +
" correct: " + total_correct + "<br>" +
" score: " + greenspan + percent.toFixed(2) + "</span>";
AB.msg ( thehtml, 6 );
test_index++;
if ( test_index == NOTEST )
{
console.log( "finished testrun: " + testrun + " score: " + percent.toFixed(2) );
testrun++;
test_index = 0;
total_tests = 0;
total_correct = 0;
}
}
//--- find no.1 (and maybe no.2) output nodes ---------------------------------------
// (restriction) assumes array values start at 0 (which is true for output nodes)
function find12 (a) // return array showing indexes of no.1 and no.2 values in array
{
let no1 = 0;
let no2 = 0;
let no1value = 0;
let no2value = 0;
for (let i = 0; i < a.length; i++)
{
if (a[i] > no1value) // new no1
{
// old no1 becomes no2
no2 = no1;
no2value = no1value;
// now put in the new no1
no1 = i;
no1value = a[i];
}
else if (a[i] > no2value) // new no2
{
no2 = i;
no2value = a[i];
}
}
var b = [ no1, no2 ];
return b;
}
// just get the maximum - separate function for speed - done many times
// find our guess - the max of the output nodes array
function findMax (a)
{
let no1 = 0;
let no1value = 0;
for (let i = 0; i < a.length; i++)
{
if (a[i] > no1value)
{
no1 = i;
no1value = a[i];
}
}
return no1;
}
// --- the draw function -------------------------------------------------------------
// every step:
function draw()
{
// check if libraries and data loaded yet:
if ( typeof mnist == 'undefined' ) return;
// how can we get white doodle on black background on yellow canvas?
// background('#ffffcc'); doodle.background('black');
background ('black');
if ( do_training )
{
// do some training per step
for (let i = 0; i < TRAINPERSTEP; i++)
{
if (i == 0) trainit(true); // show only one per step - still flashes by
else trainit(false);
}
// do some testing per step
for (let i = 0; i < TESTPERSTEP; i++)
testit();
}
// keep drawing demo and doodle images
// and keep guessing - we will update our guess as time goes on
if ( demo_exists )
{
drawDemo();
guessDemo();
}
if ( doodle_exists )
{
drawDoodle();
guessDoodle();
}
// detect doodle drawing
// (restriction) the following assumes doodle starts at 0,0
if ( mouseIsPressed ) // gets called when we click buttons, as well as if in doodle corner
{
// console.log ( mouseX + " " + mouseY + " " + pmouseX + " " + pmouseY );
var MAX = ZOOMPIXELS + 20; // can draw up to this pixels in corner
if ( (mouseX < MAX) && (mouseY < MAX) && (pmouseX < MAX) && (pmouseY < MAX) )
{
mousedrag = true; // start a mouse drag
doodle_exists = true;
doodle.stroke('white');
doodle.strokeWeight( DOODLE_THICK );
doodle.line(mouseX, mouseY, pmouseX, pmouseY);
}
}
else
{
// are we exiting a drawing
if ( mousedrag )
{
mousedrag = false;
// console.log ("Exiting draw. Now blurring.");
doodle.filter (BLUR, DOODLE_BLUR); // just blur once
// console.log (doodle);
}
}
}
//--- demo -------------------------------------------------------------
// demo some test image and predict it
// get it from test set so have not used it in training
function makeDemo()
{
demo_exists = true;
var i = AB.randomIntAtoB ( 0, NOTEST - 1 );
demo = mnist.test_images[i];
var label = mnist.test_labels[i];
thehtml = "Test image no: " + i + "<br>" +
"Classification: " + label + "<br>" ;
AB.msg ( thehtml, 8 );
// type "demo" in console to see raw data
}
function drawDemo()
{
var theimage = getImage ( demo );
// console.log (theimage);
image ( theimage, 0, canvasheight - ZOOMPIXELS, ZOOMPIXELS, ZOOMPIXELS ); // magnified
image ( theimage, ZOOMPIXELS+50, canvasheight - ZOOMPIXELS, PIXELS, PIXELS ); // original
}
function guessDemo()
{
let inputs = getInputs ( demo );
demo_inputs = inputs; // can inspect in console
let prediction = nn.predict(inputs); // array of outputs
let guess = findMax(prediction); // the top output
thehtml = " We classify it as: " + greenspan + guess + "</span>" ;
AB.msg ( thehtml, 9 );
}
//--- doodle -------------------------------------------------------------
function drawDoodle()
{
// doodle is createGraphics not createImage
let theimage = doodle.get();
// console.log (theimage);
image ( theimage, 0, 0, ZOOMPIXELS, ZOOMPIXELS ); // original
image ( theimage, ZOOMPIXELS+50, 0, PIXELS, PIXELS ); // shrunk
}
function guessDoodle()
{
// doodle is createGraphics not createImage
let img = doodle.get();
img.resize ( PIXELS, PIXELS );
img.loadPixels();
// set up inputs
let inputs = [];
for (let i = 0; i < PIXELSSQUARED ; i++)
{
inputs[i] = img.pixels[i * 4] / 255;
}
doodle_inputs = inputs; // can inspect in console
// feed forward to make prediction
let prediction = nn.predict(inputs); // array of outputs
let b = find12(prediction); // get no.1 and no.2 guesses
thehtml = " We classify it as: " + greenspan + b[0] + "</span> <br>" +
" No.2 guess is: " + greenspan + b[1] + "</span>";
AB.msg ( thehtml, 2 );
}
function wipeDoodle()
{
doodle_exists = false;
doodle.background('black');
}
// --- debugging --------------------------------------------------
// in console
showInputs(demo_inputs);
showInputs(doodle_inputs);
function showInputs ( inputs )
// display inputs row by row, corresponding to square of pixels
{
var str = "";
for (let i = 0; i < inputs.length; i++)
{
if ( i % PIXELS == 0 ) str = str + "\n"; // new line for each row of pixels
var value = inputs[i];
str = str + " " + value.toFixed(2) ;
}
console.log (str);
}
window.onload = function () {
// create a new game object which is an instance of Phaser.Game
var game = new Phaser.Game(1280, 720, Phaser.CANVAS);
// add all States to the game object (this program has only the Main State)
game.state.add('MainState', App.MainState);
// start the Main State
game.state.start('MainState');
};
/***********************************************************************************
/* The Application Namespace
/***********************************************************************************/
var App = App || {};
// ---------------------------------------------------------------------------------
// Global constants and variables
// ---------------------------------------------------------------------------------
// the names of all datasets
App.DATASETS = ['car', 'fish', 'snowman'];
// ---------------------------------------------------------------------------------
// The Main State constructor
// ---------------------------------------------------------------------------------
App.MainState = function(){
// constants describing all modes of the main state
this.MODE_INIT = 1;
this.MODE_OPEN_FILE = 2;
this.MODE_LOAD_FILE = 3;
this.MODE_START_TRAIN = 4;
this.MODE_DO_TRAIN = 5;
this.MODE_START_PREDICT = 6;
this.MODE_DO_PREDICT = 7;
this.MODE_DRAW = 8;
// set initial mode
this.mode = this.MODE_INIT;
// the counter of currently loaded datasets
this.dataset = 0;
};
// ---------------------------------------------------------------------------------
// The Main State prototype
// ---------------------------------------------------------------------------------
App.MainState.prototype = {
/**
* Automatically called only once to load all assets.
*/
preload : function(){
this.game.load.image('imgBack', '../assets/img_back_2.png');
this.game.load.image('btnMoreGames', '../assets/btn_moregames.png');
this.game.load.image('btnAuthor', '../assets/btn_author.png');
this.load.bitmapFont('fntBlackChars', '../assets/fnt_black_chars.png', '../assets/fnt_black_chars.fnt');
},
/**
* Automatically called immediately after all assets are loaded to create all objects.
*/
create : function(){
// scale game to cover the entire screen
this.scale.scaleMode = Phaser.ScaleManager.SHOW_ALL;
this.scale.pageAlignVertically = true;
this.scale.pageAlignHorizontally = true;
// keep game running if it loses the focus
this.game.stage.disableVisibilityChange = true;
// create background
this.game.add.sprite(0, 0, 'imgBack');
// create a loader for loading datasets
this.loader = new Phaser.Loader(this.game);
// create user interface with buttons, bitmaps and texts
this.ui = new UI(this);
// create a convolution neural network
this.cnn = new CNN(this);
},
/**
* Automatically called on every tick representing the main loop.
*/
update : function(){
switch(this.mode){
// initialize the game
case this.MODE_INIT:
this.ui.disableButtons();
this.mode = this.MODE_OPEN_FILE;
break;
// open dataset file and start loading it
case this.MODE_OPEN_FILE:
var fileName = App.DATASETS[this.dataset] + '.bin';
this.loader.reset();
this.loader.binary('input_file', '../data/'+fileName);
this.loader.start();
this.ui.showStatusBar("Loading " + fileName + " file.");
this.mode = this.MODE_LOAD_FILE;
break;
// wait on dataset file to be loaded
case this.MODE_LOAD_FILE:
if (this.loader.hasLoaded){
// split the loaded dataset into training data and test data
this.cnn.splitDataset(
new Uint8Array(this.game.cache.getBinary('input_file')),
this.dataset
);
// increase the number of loaded datasets
this.dataset++;
// if we have not loaded all datasets yet then go to load the next one
if (this.dataset < App.DATASETS.length){
this.mode = this.MODE_OPEN_FILE;
} else {
this.ui.showStatusBar("All datasets loaded.");
this.mode = this.MODE_START_TRAIN;
}
}
break;
case this.MODE_START_TRAIN:
this.mode = this.MODE_DO_TRAIN;
break;
case this.MODE_DO_TRAIN:
this.mode = this.MODE_START_PREDICT;
break;
case this.MODE_START_PREDICT:
this.mode = this.MODE_DO_PREDICT;
break;
case this.MODE_DO_PREDICT:
this.mode = this.MODE_DRAW;
break;
case this.MODE_DRAW:
break;
}
}
};
//The CNN Classifier
var CNN = function(main){
// reference to the Main State
this.main = main;
this.NUM_CLASSES = App.DATASETS.length; // number of classes which can be recognized by CNN model
this.IMAGE_SIZE = 784; // size of an image in a dataset
this.NUM_TRAIN_IMAGES = 400; // number of training images in a dataset
this.NUM_TEST_IMAGES = 100; // number of test images in a dataset
// total number of training images in all classes
const TOTAL_TRAIN_IMAGES = this.NUM_CLASSES * this.NUM_TRAIN_IMAGES;
// total number of test images in all classes
const TOTAL_TEST_IMAGES = this.NUM_CLASSES * this.NUM_TEST_IMAGES;
// create Training Data arrays for storing training images and their corresponding classes
this.aTrainImages = new Float32Array(TOTAL_TRAIN_IMAGES * this.IMAGE_SIZE);
this.aTrainClasses = new Uint8Array(TOTAL_TRAIN_IMAGES);
// shuffle Training Data by creating an array of shuffled Train indices
this.aTrainIndices = tf.util.createShuffledIndices(TOTAL_TRAIN_IMAGES);
// the reference to the current element in the aTrainIndices[] array
this.trainElement = -1;
// create arrays of Test Data for storing test images and their corresponding classes
this.aTestImages = new Float32Array(TOTAL_TEST_IMAGES * this.IMAGE_SIZE);
this.aTestClasses = new Uint8Array(TOTAL_TEST_IMAGES);
// shuffle Test Data by creating an array of shuffled Test indices
this.aTestIndices = tf.util.createShuffledIndices(TOTAL_TEST_IMAGES);
// the reference to the current element in the aTestIndices[] array
this.testElement = -1;
};
// ---------------------------------------------------------------------------------
// CNN Prototype
// ---------------------------------------------------------------------------------
/**
* Splits the entire dataset into training data and test data.
*
* @param {Uint8Array} imagesBuffer - the array with binary data of all images in the dataset
* @param {integer} dataset - the ordinal number of the dataset
*/
CNN.prototype.splitDataset = function(imagesBuffer, dataset){
// slice dataset to get training images and normalize them
var trainBuffer = new Float32Array(imagesBuffer.slice(0, this.IMAGE_SIZE * this.NUM_TRAIN_IMAGES));
trainBuffer = trainBuffer.map(function (cv){return cv/255.0});
// add training images and their corresponding classes into Training Data arrays
var start = dataset * this.NUM_TRAIN_IMAGES;
this.aTrainImages.set(trainBuffer, start * this.IMAGE_SIZE);
this.aTrainClasses.fill(dataset, start, start + this.NUM_TRAIN_IMAGES);
// slice dataset to get test images and normalize them
var testBuffer = new Float32Array(imagesBuffer.slice(this.IMAGE_SIZE * this.NUM_TRAIN_IMAGES));
testBuffer = testBuffer.map(function (cv){return cv/255.0});
// add test images and their corresponding classes into Test Data arrays
start = dataset * this.NUM_TEST_IMAGES;
this.aTestImages.set(testBuffer, start * this.IMAGE_SIZE);
this.aTestClasses.fill(dataset, start, start + this.NUM_TEST_IMAGES);
};
//Training the CNN Classifier
var CNN = function(main){
// reference to the Main State
this.main = main;
this.NUM_CLASSES = App.DATASETS.length; // number of classes which can be recognized by CNN model
this.IMAGE_SIZE = 784; // size of an image in a dataset
this.NUM_TRAIN_IMAGES = 400; // number of training images in a dataset
this.NUM_TEST_IMAGES = 100; // number of test images in a dataset
this.TRAIN_ITERATIONS = 50; // total number of training iterations
this.TRAIN_BATCH_SIZE = 100; // number of training images used to train model during one iteration
this.TEST_FREQUENCY = 5; // frequency of testing model accuracy (one test on every 5 training iterations)
this.TEST_BATCH_SIZE = 50; // number of test images used to test model accuracy
this.trainIteration = 0; // current number of executed training iterations
this.aLoss = []; // array to store model's loss values during training
this.aAccu = []; // array to store model's accuracy values during training
// total number of training images in all classes
const TOTAL_TRAIN_IMAGES = this.NUM_CLASSES * this.NUM_TRAIN_IMAGES;
// total number of test images in all classes
const TOTAL_TEST_IMAGES = this.NUM_CLASSES * this.NUM_TEST_IMAGES;
// create Training Data arrays for storing training images and their corresponding classes
this.aTrainImages = new Float32Array(TOTAL_TRAIN_IMAGES * this.IMAGE_SIZE);
this.aTrainClasses = new Uint8Array(TOTAL_TRAIN_IMAGES);
// shuffle Training Data by creating an array of shuffled Train indices
this.aTrainIndices = tf.util.createShuffledIndices(TOTAL_TRAIN_IMAGES);
// the reference to the current element in the aTrainIndices[] array
this.trainElement = -1;
// create arrays of Test Data for storing test images and their corresponding classes
this.aTestImages = new Float32Array(TOTAL_TEST_IMAGES * this.IMAGE_SIZE);
this.aTestClasses = new Uint8Array(TOTAL_TEST_IMAGES);
// shuffle Test Data by creating an array of shuffled Test indices
this.aTestIndices = tf.util.createShuffledIndices(TOTAL_TEST_IMAGES);
// the reference to the current element in the aTestIndices[] array
this.testElement = -1;
// create a CNN model using a Sequential model type in which
// tensors are consecutively passed from one layer to the next
this.model = tf.sequential();
// add a convolutional layer
this.model.add(tf.layers.conv2d({
inputShape: [28, 28, 1],
kernelSize: 5,
filters: 8,
strides: 1,
activation: 'relu',
kernelInitializer: 'varianceScaling'
}));
// add a max pooling layer
this.model.add(tf.layers.maxPooling2d({
poolSize: [2, 2],
strides: [2, 2]
}));
// add a second convolutional layer
this.model.add(tf.layers.conv2d({
kernelSize: 5,
filters: 16,
strides: 1,
activation: 'relu',
kernelInitializer: 'varianceScaling'
}));
// add a second max pooling layer
this.model.add(tf.layers.maxPooling2d({
poolSize: [2, 2],
strides: [2, 2]
}));
// add a flatten layer to flatten the output of the previous layer to a vector
this.model.add(tf.layers.flatten());
// add a dense (fully connected) layer to perform the final classification
this.model.add(tf.layers.dense({
units: this.NUM_CLASSES,
kernelInitializer: 'varianceScaling',
activation: 'softmax'
}));
// compile the model
this.model.compile({
optimizer: tf.train.sgd(0.15), // optimizer with learning rate
loss: 'categoricalCrossentropy', // loss function
metrics: ['accuracy'], // evaluation metric
});
};
// ---------------------------------------------------------------------------------
// CNN Prototype
// ---------------------------------------------------------------------------------
/**
* Splits the entire dataset into training data and test data.
*
* @param {Uint8Array} imagesBuffer - the array with binary data of all images in the dataset
* @param {integer} dataset - the ordinal number of the dataset
*/
CNN.prototype.splitDataset = function(imagesBuffer, dataset){
// slice dataset to get training images and normalize them
var trainBuffer = new Float32Array(imagesBuffer.slice(0, this.IMAGE_SIZE * this.NUM_TRAIN_IMAGES));
trainBuffer = trainBuffer.map(function (cv){return cv/255.0});
// add training images and their corresponding classes into Training Data arrays
var start = dataset * this.NUM_TRAIN_IMAGES;
this.aTrainImages.set(trainBuffer, start * this.IMAGE_SIZE);
this.aTrainClasses.fill(dataset, start, start + this.NUM_TRAIN_IMAGES);
// slice dataset to get test images and normalize them
var testBuffer = new Float32Array(imagesBuffer.slice(this.IMAGE_SIZE * this.NUM_TRAIN_IMAGES));
testBuffer = testBuffer.map(function (cv){return cv/255.0});
// add test images and their corresponding classes into Test Data arrays
start = dataset * this.NUM_TEST_IMAGES;
this.aTestImages.set(testBuffer, start * this.IMAGE_SIZE);
this.aTestClasses.fill(dataset, start, start + this.NUM_TEST_IMAGES);
};
/**
* Trains the model
*/
CNN.prototype.train = async function(){
// reset the training flag to know the training is currently in progress
this.isTrainCompleted = false;
for (let item = 0; item < this.TRAIN_ITERATIONS; item++){
// increase the number of training iterations
this.trainIteration++;
this.main.ui.showStatusBar("Training the CNN - iteration " + this.trainIteration + ".");
// fetch the next Training Batch
let trainBatch = this.nextTrainBatch(this.TRAIN_BATCH_SIZE);
// create new Test Batch and Validation Set
let testBatch;
let validationSet;
if (item % this.TEST_FREQUENCY === 0){ // every few training passes...
// fetch the next Test Batch
testBatch = this.nextTestBatch(this.TEST_BATCH_SIZE);
// build Validation Set by using images and corresponding labels from Test Batch
validationSet = [testBatch.images, testBatch.labels];
}
// train the model
const training = await this.model.fit(
trainBatch.images,
trainBatch.labels,
{batchSize: this.TRAIN_BATCH_SIZE, validationData: validationSet, epochs: 1}
);
// get the model loss from the last training iteration and plot the Loss Chart
var maxLossLength = this.main.ui.bmpLossChart.width;
if (this.aLoss.length > maxLossLength) this.aLoss.shift();
this.aLoss.push(1 - Math.min(1, training.history.loss[0]));
this.main.ui.plotChart(this.main.ui.bmpLossChart, this.aLoss, 1);
if (testBatch != null) {
// get the model accuracy from the last training iteration and plot the Accuracy Chart
var maxAccuLength = this.main.ui.bmpAccuChart.width;
if (this.aAccu.length * this.TEST_FREQUENCY > maxAccuLength) this.aAccu.shift();
this.aAccu.push(1 - Math.min(1, training.history.acc[0]));
this.main.ui.plotChart(this.main.ui.bmpAccuChart, this.aAccu, this.TEST_FREQUENCY);
// dispose Test Batch from memory
testBatch.images.dispose();
testBatch.labels.dispose();
}
// dispose Training Batch from memory
trainBatch.images.dispose();
trainBatch.labels.dispose();
// mitigate blocking the UI thread and freezing the tab during training
await tf.nextFrame();
}
// set the training flag to know the training is completed
this.isTrainCompleted = true;
};
/**
* Returns a batch of images and their corresponding classes from the Training Data
*
* @param {integer} batchSize - how many images are included in training batch
*/
CNN.prototype.nextTrainBatch = function(batchSize){
return this.fetchBatch(
batchSize, this.aTrainImages, this.aTrainClasses,
() => {
this.trainElement = (this.trainElement + 1) % this.aTrainIndices.length;
return this.aTrainIndices[this.trainElement];
}
);
};
/**
* Returns a batch of images and their corresponding classes from the Test Data
*
* @param {integer} batchSize - how many images are included in test batch
*/
CNN.prototype.nextTestBatch = function(batchSize){
return this.fetchBatch(
batchSize, this.aTestImages, this.aTestClasses,
() => {
this.testElement = (this.testElement + 1) % this.aTestIndices.length;
return this.aTestIndices[this.testElement];
}
);
};
/**
* Fetches a batch of images and their corresponding classes
*
* @param {integer} batchSize - how many images are included in the batch
* @param {Float32Array} aImages - array of images
* @param {Uint8Array} aClasses - array of corresponding classes
* @param {integer} getIndex - a function which returns the index of an image that must be fetched
*/
CNN.prototype.fetchBatch = function(batchSize, aImages, aClasses, getIndex){
// create batch arrays
const batchImages = new Float32Array(batchSize * this.IMAGE_SIZE);
const batchLabels = new Uint8Array(batchSize * this.NUM_CLASSES);
for (let item = 0; item < batchSize; item++){
// get the index of the image we want to fetch
const idx = getIndex();
// fetch the image
const image = aImages.slice(idx * this.IMAGE_SIZE, (idx + 1) * this.IMAGE_SIZE);
// add the image to the batch of images
batchImages.set(image, i * this.IMAGE_SIZE);
// get the class number of this image
const class_num = aClasses[idx];
// generate the label for this image by using "one hot encoding method":
// define a vector where all elements are 0, beside one element
// which points to the correct class of this image
const label = new Uint8Array(this.NUM_CLASSES);
label[class_num] = 1;
// add the label to the batch of labels
batchLabels.set(label, i * this.NUM_CLASSES);
}
// convert batch of images to a temporary tensor
const images_temp = tf.tensor2d(batchImages, [batchSize, this.IMAGE_SIZE]);
// reshape the temporary tensor to the size of the model input shape
const images = images_temp.reshape([batchSize, 28, 28, 1]);
// dispose the temporary tensor to free memory
images_temp.dispose();
// convert batch of labels to tensor
const labels = tf.tensor2d(batchLabels, [batchSize, this.NUM_CLASSES]);
return {images, labels};
};