// Cloned by Deniss Strods on 3 Dec 2020 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; // 24873;//60000;
const NOTEST = 10000;
//--- can modify all these --------------------------------------------------
// no of nodes in network
const noinput = PIXELSSQUARED;
const nohidden = 200;
const nooutput = 10;
const learningrate = 0.01; // default 0.1
let decayLearningRate = learningrate;
// should we train every timestep or not
let do_training = false;
// how many to train and test per timestep
const TRAINPERSTEP = 20;
const TESTPERSTEP = 3;
// multiply it by this to magnify for display
const ZOOMFACTOR = 10;
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 //18 originally
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 percent;
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
}
// CSS trick
// make run header bigger
// $("#runheaderbox").css({
// "max-height": "95vh",
// });
$('#ab-runheaderbox').attr('style', 'position: initial !important; max-height:95vh; margin-left: 350px');
//--- 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> " +
`<div><h3>Processed image</h3><canvas id="can" width="280" height="280"></canvas></div>`
AB.msg(thehtml, 1);
// 2 Doodle variable data (guess)
// 3 Training header
// Generating table with good models, Author Deniss Strods
const createTableOfSavedModels = (savedModels, modelLoaded) => {
let table =
`<h3>Saved Models ${modelLoaded ? '<span style="color:green;">(Model loaded)</span>' : ''}<h3><table>`;
table += '<tr><th>Action</th><th>Date</th><th>Acuracy</th><th>Comment</th></tr>';
if (savedModels) {
savedModels.forEach((model, index) => {
const button =
`<button onclick='loadDataIntoModel(savedItems[${index}]);' class='normbutton' >Load</button>`;
table +=
`<tr><td>${button}</td><td>${model.date}</td><td>${model.lastPercent && model.lastPercent.toFixed(2)}</td><td>${model.comment ? model.comment : 'other user created' }</td></tr>`;
})
}
table += '</table>';
return table;
}
// Refreshing saved table, Author Deniss Strods
let savedItems;
const refreshItems = (savedItems, modelLoaded = false) => {
let thehtml =
"<hr> <h1> 2. Training </h1> Middle row: Training image magnified (left) and original (right). <br> " +
" <button onclick='do_training = !do_training;' class='normbutton' >Stop / Start training</button><br> <button hidden onclick='saveCurrentWeights()' class='normbutton' >Save model</button> <br>";
//<button onclick='saveCurrentWeights()' class='normbutton' >Save model</button> //removing this button so good model would not be owerwritten
thehtml += createTableOfSavedModels(savedItems, modelLoaded);
AB.msg(thehtml, 4);
}
refreshItems();
// Pulling out saved items, Author Deniss Strods
const getSavedItems = () => {
AB.getAllData(objects => {
console.log("Getting Saved Items", objects);
const items = objects.map( o => o[2]);
const elements = [];
items.forEach(i=> {
elements.push(...i);
})
savedItems = elements;
refreshItems(savedItems);
})
}
// Save current weights, Author Deniss Strods
const saveCurrentWeights = () => {
if (AB.runloggedin) {
const newItem = {
date: new Date(),
layers: nn.layers,
lastPercent: percent,
comment: 'Mark, use this for test'
};
savedItems = [newItem];
console.log("Saving Data:", savedItems);
AB.saveData(savedItems);
refreshItems(savedItems);
}
}
// Loading data into the model, Author Deniss Strods
const loadDataIntoModel = (model) => {
do_training = false;
//allow time to finish all the operations
setTimeout(() => {
nn.loadModel(model);
refreshItems(savedItems, true);
}, 500);
//Reset all the params
trainrun = 1;
train_index = 0;
testrun = 1;
test_index = 0;
total_tests = 0;
total_correct = 0;
}
// 4 variable training data
// 5 Testing header
thehtml = "<h3> Hidden tests </h3> ";
AB.msg(thehtml, 6);
// 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, 8);
// 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() {
getSavedItems();
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();
});
// });
});
wipeDoodle();
}
let specificIndexes = [];
// 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
// Selective training, pick what additional numbers to train upon, Author Deniss Strods...
// let img = mnist.train_images[train_index];
// let label = mnist.train_labels[train_index];
// mnist.train_labels.forEach( (l,index)=> {
// if (l == 9 || l == 6 || l == 7 || l == 1){
// specificIndexes.push(index);
// }
// });
// console.log("INDEXES...", specificIndexes.length);
});
}
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
{
// const index = specificIndexes[train_index];
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
train_inputs = inputs; // can inspect in console
nn.train(inputs, targets);
thehtml = " trainrun: " + trainrun + "<br> no: " + train_index + ` / learning rate: ${decayLearningRate}`;
AB.msg(thehtml, 5);
train_index++;
if (train_index == NOTRAIN) {
train_index = 0;
console.log("finished trainrun: " + trainrun);
trainrun++;
// learning decay added
decayLearningRate = stepDecay(trainrun, learningrate);
nn.setLearningRate(decayLearningRate);
console.log("setting learning rate: ", decayLearningRate);
}
}
// Investigated dynamic learning rate and identified that there may be benifit in adding learning rate decay
// https://towardsdatascience.com/learning-rate-schedules-and-adaptive-learning-rate-methods-for-deep-learning-2c8f433990d1
// Author Deniss Strods
const stepDecay = (epoch, initialRate) => {
const drop = 0.5
const epochsDrop = 10
const lrate = initialRate * Math.pow(drop,
Math.floor((1 + epoch) / epochsDrop))
return lrate
}
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++;
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) {
no1 = i;
no1value = a[i];
} else if (a[i] > no2value) {
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 perdictDoodle() {
guessDoodle();
}
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;
guessDoodle();
}
}
}
//--- 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, 9);
// 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
showInputs(demo_inputs);
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, 10);
}
//--- 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.loadPixels();
// Processing the doodle image
img = getCanvasImagePixels(img);
// 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, 3);
}
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);
}
// Other techniques for learning
class ActivationFunction {
constructor(func, dfunc) {
this.func = func;
this.dfunc = dfunc;
}
}
let sigmoid = new ActivationFunction(
x => 1 / (1 + Math.exp(-x)),
y => y * (1 - y)
);
let tanh = new ActivationFunction(
x => Math.tanh(x),
y => 1 - (y * y)
);
const softmax = (arr) => {
const C = Math.max(...arr);
const d = arr.map((y) => Math.exp(y - C)).reduce((a, b) => a + b);
return arr.map((value, index) => {
return Math.exp(value - C) / d;
})
}
const softmaxJacobian = (arr) => {
let jacobian = [...new Array(arr.length)].map(x => new Array(arr.length));
for (let i = 0; i < arr.length; i++) {
for (let j = 0; j < arr.length; j++) {
if (i == j) {
jacobian[i][j] = arr[i] * (1 - arr[i]);
} else {
jacobian[i][j] = -arr[i] * arr[j];
}
}
}
return jacobian;
}
// We are not using error function in softmax deriviative, Author Deniss Strods
const errorFunction = (y, o) => -y * Math.log(o); // y expected correct value, o actual output value
// Softmax deriviative, Author Deniss Strods
const dSoftmax = (y, o) => o - y; // y expected correct value, o actual output value
// NN LAyer representation class, author Deniss Strods
class NeuralNetworkLayer {
constructor(in_nodes, hid_nodes, activationFunction) {
this.input_nodes = in_nodes;
this.hidden_nodes = hid_nodes;
this.weights_ih = new Matrix(this.hidden_nodes, this
.input_nodes);
this.weights_ih.randomize();
this.bias_h = new Matrix(this.hidden_nodes, 1);
this.bias_h.randomize();
this.activationFunction = activationFunction;
}
}
// HEavily modified by Deniss Strods, almost nothing left from original,
class NeuralNetwork {
constructor(in_nodes, hid_nodes, out_nodes) {
// adding softmax layer and 2 hidden layers, selecting activation function for layers
const hiddenLayer = new NeuralNetworkLayer(in_nodes,
hid_nodes, 'sigmoid');
const secondHiddenLayer = new NeuralNetworkLayer(hid_nodes,
hid_nodes, 'sigmoid');
const outputLayer = new NeuralNetworkLayer(hid_nodes, out_nodes, 'softmax');
this.layers = [hiddenLayer, secondHiddenLayer, outputLayer];
this.setLearningRate();
}
// passing trough the layers in the layers list, calculaing and getting perdiction
forwardPass(inputs, layers) {
// Iterating trough all the layers forward
let previousLayer;
// allowing to set up differen activation function per layer
const calculatedLayers = layers.map((layer, index) => {
let matrix = Matrix.multiply(layer.weights_ih,
!previousLayer ? inputs : previousLayer
);
matrix.add(layer.bias_h);
// adding ability to identify
if (layer.activationFunction === 'softmax') {
const softmaxData = softmax(matrix.data).map(e => [e]);
matrix.data = softmaxData;
} else if (layer.activationFunction === 'sigmoid') {
matrix.map(sigmoid.func);
} else if (layer.activationFunction === 'tanh') {
matrix.map(tanh.func);
}
previousLayer = matrix;
return matrix;
})
return calculatedLayers;
}
predict(input_array) {
// // Generating the Hidden Outputs
let inputs = Matrix.fromArray(input_array);
const calculatedLayers = this.forwardPass(inputs, this.layers)
// Sending back to the caller!
return calculatedLayers[calculatedLayers.length - 1].toArray();
}
setLearningRate(learning_rate = 0.1) {
this.learning_rate = learning_rate;
}
setActivationFunction(func = sigmoid) {
this.activation_function = func;
}
train(input_array, target_array) {
// Generating the Hidden Outputs
let inputs = Matrix.fromArray(input_array);
const calculatedLayers = this.forwardPass(inputs, this.layers)
// This is backwards pass
let error;
for (let i = calculatedLayers.length - 1; i >= 0; i--) {
const layer = calculatedLayers[i];
const previousLayer = (i - 1) < 0 ? inputs :
calculatedLayers[i - 1];
let e;
if (!error) {
e = Matrix.subtract(Matrix.fromArray(
target_array), layer);
} else {
const nextLayer = this.layers[i + 1];
e = Matrix.multiply(Matrix.transpose(nextLayer
.weights_ih), error);
}
error = e;
let gradients;
// allowing to set up differen activation function per layer
if (this.layers[i].activationFunction === 'softmax') {
let errorSumm = 0;
const gSoftmax = target_array.map((item, index) => {
return [dSoftmax(item, layer.data[index]) * -1];
});
const softmaxGradients = new Matrix(10, 1);
softmaxGradients.data = gSoftmax;
gradients = softmaxGradients;
} else if (this.layers[i].activationFunction === 'sigmoid') {
gradients = Matrix.map(layer,
sigmoid.dfunc);
gradients.multiply(e);
} else if (this.layers[i].activationFunction === 'tanh') {
gradients = Matrix.map(layer,
tanh.dfunc);
gradients.multiply(e);
}
gradients.multiply(this.learning_rate);
// Calculate deltas
let weight_ho_deltas = Matrix.multiply(gradients,
Matrix.transpose(previousLayer));
// Adjust the weights by deltas
this.layers[i].weights_ih.add(weight_ho_deltas);
// Adjust the bias by its deltas (which is just the gradients)
this.layers[i].bias_h.add(gradients);
}
}
// Accept an arbitrary function for mutation
mutate(func) {
this.weights_ih.map(func);
this.weights_ho.map(func);
this.bias_h.map(func);
this.bias_o.map(func);
}
remapMatricFromObject(matrix, object) {
const {
data,
rows,
cols
} = object;
matrix.data = data;
matrix.rows = rows;
matrix.cols = cols;
}
// Loading model, Author Deniss Strods
loadModel(model) {
const {
layers
} = model
if (this.layers.length === layers.length) {
console.log("Loading model", model);
this.layers.forEach((layer, index) => {
const {
weights_ih,
bias_h,
input_nodes,
hidden_nodes,
activationFunction
} = layers[index];
this.remapMatricFromObject(layer.weights_ih, weights_ih);
this.remapMatricFromObject(layer.bias_h, bias_h);
layer.activationFunction = activationFunction;
layer.input_nodes = input_nodes;
layer.hidden_nodes = hidden_nodes;
});
}
}
}
/// this was takken from myselph.de/neuralNet.html
// given grayscale image, find bounding rectangle of digit defined
// by above-threshold surrounding
function getBoundingRectangle(img, threshold) {
var rows = img.length;
var columns = img[0].length;
var minX = columns;
var minY = rows;
var maxX = -1;
var maxY = -1;
for (var y = 0; y < rows; y++) {
for (var x = 0; x < columns; x++) {
if (img[y][x] > threshold) {
if (minX > x) minX = x;
if (maxX < x) maxX = x;
if (minY > y) minY = y;
if (maxY < y) maxY = y;
}
}
}
return {
minY: minY,
minX: minX,
maxY: maxY,
maxX: maxX
};
}
/// this was takken from myselph.de/neuralNet.html
// take canvas image and convert to grayscale. Mainly because my
// own functions operate easier on grayscale, but some stuff like
// resizing and translating is better done with the canvas functions
function imageDataToGrayscale(imgData) {
const grayscaleImg = [];
for (var y = 0; y < imgData.height; y++) {
grayscaleImg[y] = [];
for (var x = 0; x < imgData.width; x++) {
var offset = y * 4 * imgData.width + 4 * x;
var alpha = imgData.data[offset + 3];
// weird: when painting with stroke, alpha == 0 means white;
// alpha > 0 is a grayscale value; in that case I simply take the R value
if (alpha == 0) {
imgData.data[offset] = 255;
imgData.data[offset + 1] = 255;
imgData.data[offset + 2] = 255;
}
imgData.data[offset + 3] = 255;
// simply take red channel value. Not correct, but works for
// black or white images.
grayscaleImg[y][x] = imgData.data[y * 4 * imgData.width + x * 4 + 0] / 255;
}
}
return grayscaleImg;
}
// example input {minY: 44, minX: 137, maxY: 249, maxX: 162}
// finding necesarry shift based on the min max of X and Y author Deniss Strods
const findShiftAndScale = ({
minY,
minX,
maxY,
maxX
}) => {
const subSquareSize = 200;
const squareSize = 280;
const objHeight = maxY - minY;
const objWidth = maxX - minX;
const idealGapY = Math.floor((squareSize - objHeight) / 2);
const idealGapX = Math.floor((squareSize - objWidth) / 2);
// shift needed to center
const minYdelta = idealGapY - minY;
const minXdelta = idealGapX - minX;
// scale needed
const ratioY = subSquareSize / objWidth;
const ratioX = subSquareSize / objHeight;
const sortedList = [ratioY, ratioX].sort();
return {
minYdelta,
minXdelta,
scale: sortedList[0]
}
}
// Shifting context, author Deniss Strods
const shiftContext = (ctx, w, h, dx, dy) => {
const imageData = ctx.getImageData(0, 0, w, h);
ctx.strokeStyle = "black";
ctx.fillStyle = "black";
ctx.fillRect(0, 0, w, h);
ctx.putImageData(imageData, dx, dy);
}
// Resizing image based on ratio, author Deniss Strods
const resizeTo = (canvas, pct) => {
const tempCanvas = document.createElement("canvas");
const tctx = tempCanvas.getContext("2d");
const cw = canvas.width;
const ch = canvas.height;
tempCanvas.width = cw;
tempCanvas.height = ch;
tctx.setTransform(pct, 0, 0, pct, tctx.canvas.width / 2, tctx.canvas.height / 2);
tctx.drawImage(canvas, -canvas.width / 2, -canvas.height / 2); // draw the image offset by half
const ctx = canvas.getContext('2d');
ctx.drawImage(tempCanvas, 0, 0);
}
// Getting pixels for comparison, author Deniss Strods
const getCanvasImagePixels = (image) => {
let {
imageData: imgData,
canvas
} = image;
let grayscaleImg = imageDataToGrayscale(imgData);
imgData = image.drawingContext.getImageData(0, 0, 280, 280);
const boundingRectangle = getBoundingRectangle(grayscaleImg, 0.02);
const canvasCopy = document.createElement("canvas");
canvasCopy.width = imgData.width;
canvasCopy.height = imgData.height;
const copyCtx = canvasCopy.getContext("2d");
copyCtx.drawImage(image.drawingContext.canvas, 0, 0);
const {
minYdelta,
minXdelta,
scale
} = findShiftAndScale(boundingRectangle)
shiftContext(copyCtx, canvasCopy.width, canvasCopy.height, minXdelta, minYdelta);
resizeTo(canvasCopy, scale);
// now bin image into 10x10 blocks (giving a 28x28 image)
// fragment takken from myselph.de/neuralNet.html
imgData = copyCtx.getImageData(0, 0, 280, 280);
grayscaleImg = imageDataToGrayscale(imgData);
let nnInput = new Array(784);
for (let y = 0; y < 28; y++) {
for (let x = 0; x < 28; x++) {
let mean = 0;
for (let v = 0; v < 10; v++) {
for (let h = 0; h < 10; h++) {
mean += grayscaleImg[y * 10 + v][x * 10 + h];
}
}
mean = (1 - mean / 100); // average and invert
nnInput[x * 28 + y] = (mean - .5) / .5;
}
}
// Drawing, fragment takken from myselph.de/neuralNet.html
const imgCanvas = document.getElementById('can');
const ctx = imgCanvas.getContext("2d");
ctx.clearRect(0, 0, imgCanvas.width, imgCanvas.height);
ctx.drawImage(canvas, 0, 0);
for (var y = 0; y < 28; y++) {
for (var x = 0; x < 28; x++) {
var block = ctx.getImageData(x * 10, y * 10, 10, 10);
var newVal = 255 * (0.5 - nnInput[x * 28 + y] / 2);
for (var i = 0; i < 4 * 10 * 10; i += 4) {
block.data[i] = newVal;
block.data[i + 1] = newVal;
block.data[i + 2] = newVal;
block.data[i + 3] = 255;
}
ctx.putImageData(block, x * 10, y * 10);
}
}
const img = createImage(canvas.width, canvas.height);
img.drawingContext.drawImage(imgCanvas, 0, 0);
img.resize(PIXELS, PIXELS);
img.loadPixels();
return img;
}