let brain;

let pressed = false;

let confidence;

let classificationResult;

let labelB = '';

let labelA = '';

let labelC = '';

let labelD = '';

let confA = 0;

let confB = 0;

let confC = 0;

let confD = 0;

​

let classifierLabel;

let training = false;

let lastTraining = false;

let font;

​

​

function preload() {

  font = loadFont('SourceCodePro-Regular.ttf');

​

}

​

function setup() {

  let canvas = createCanvas(400, 400);

  canvas.parent('sketch-holder');

  // Only when clicking on the canvas

  //canvas.mousePressed(addData);

​

  //for holding down recording. 

  canvas.mousePressed(check);

  canvas.mouseReleased(uncheck);

​

​

  // Create ethe model

  const options = {

    inputs: ['x', 'y'], // TODO: support ['x', 'y']

    outputs: ['label'], // TODO: support ['label']

    debug: true,

    task: 'classification'

  }

  brain = ml5.neuralNetwork(options);

​

  background(0);

​

}

​

function draw() {

​

  addData(); //only put in draw loop if check boolean functions used. 

​

​

  if (training !== lastTraining) { //don't continuously train!

    if (training) {

      trainModel();

    }

  }

  lastTraining = training; //update last Training.

}

​

//check if buttons are pressed

function check() {

  pressed = true;

}

​

function uncheck() {

  pressed = false;

​

}

​

​

​

// Add a data record

function addData() {

  if (pressed) {

    // Get label

    let label = classifierLabel;

​

    // Add data

    brain.addData({

      x: mouseX,

      y: mouseY

    }, {

      label

    });

​

    // Draw circle to visualize training data

    push();

    stroke(255);

    noFill();

    ellipse(mouseX, mouseY, 32);

    fill(255);

    textSize(16);

    textAlign(CENTER, CENTER);

    text(label, mouseX, mouseY);

    pop();

  }

​

}

​

// Train the model

function trainModel() {

  // ml5 will normalize data to a range between 0 and 1 for you.

  brain.normalizeData();

  // Train the model

  // Epochs: one cycle through all the training data

  brain.train({

    epochs: 20

  }, finishedTraining);

}

​

// When the model is trained

function finishedTraining() {

  brain.classify([mouseX, mouseY], gotResults);

}

​

​

​

// Got a result

function gotResults(error, results) {

  if (error) {

    // console.error(error);

    return;

  }

​

​

​

  // Show classification

  confidence = nf(results[0].confidence, 0, 2);

  classificationResult = results[0].label;

  console.log(confidence);

  updateClassification(results);

​

  // Predict again

  brain.classify([mouseX, mouseY], gotResults);

}

​

​

​

//filter array function for specific labels. 

function updateClassification(results) {

  //filter the results

  var idA = results.filter(function(e) {

    return e.label == 'A';

  });

  // console.log(idA[0].confidence);

  labelA = idA[0].label;

  confA = idA[0].confidence;

​

  var idB = results.filter(function(e) {

    return e.label == 'B';

  });

  // console.log(idB[0].confidence);

  labelB = idB[0].label;

  confB = idB[0].confidence;

​

  var idC = results.filter(function(e) {

    return e.label == 'C';

  });

  // console.log(idB[0].confidence);

  labelC = idC[0].label;

  confC = idC[0].confidence;

​

  var idD = results.filter(function(e) {

    return e.label == 'D';

  });

  // console.log(idB[0].confidence);

  labelD = idD[0].label;

  confD = idD[0].confidence;

}