// The Coding Train / Daniel Shiffman

​

let model;

let targetLabel = 'C';

​

let state = 'collection';

​

let notes = {

  C: 261.6256,

  D: 293.6648,

  E: 329.6276,

  F: 349.2282,

  G: 391.9954,

  A: 440.0000,

  B: 493.8833

}

​

let env, wave;

​

function setup() {

  createCanvas(400, 400);

​

  env = new p5.Envelope();

  env.setADSR(0.05, 0.1, 0.5, 1);

  env.setRange(1.2, 0);

​

  wave = new p5.Oscillator();

​

  wave.setType('sine');

  wave.start();

  wave.freq(440);

  wave.amp(env);

​

  let options = {

    inputs: ['x', 'y'],

    outputs: ['label'],

    task: 'classification',

    debug: 'true',

    learningRate: 0.5

  };

  model = ml5.neuralNetwork(options);

  model.loadData('mouse-notes.json', dataLoaded);

​

​

  background(255);

}

​

function dataLoaded() {

  console.log(model.data);

  let data = model.data.data.raw;

  // let data = model.getData();

  for (let i = 0; i < data.length; i++) {

    let inputs = data[i].xs;

    let target = data[i].ys;

    stroke(0);

    noFill();

    ellipse(inputs.x, inputs.y, 24);

    fill(0);

    noStroke();

    textAlign(CENTER, CENTER);

    text(target.label, inputs.x, inputs.y);

  }

  state = 'training';

  console.log('starting training');

  model.normalizeData();

  let options = {

    epochs: 200

  }

  model.train(options, whileTraining, finishedTraining);

​

​

​

}

​

function keyPressed() {

​

  if (key == 't') {

    state = 'training';

    console.log('starting training');

    model.normalizeData();

    let options = {

      epochs: 200

    }

    model.train(options, whileTraining, finishedTraining);

  } else if (key == 's') {

    model.saveData('mouse-notes');

  } else {

    targetLabel = key.toUpperCase();

  }

}

​

function whileTraining(epoch, loss) {

  console.log(epoch);

}

​

function finishedTraining() {

  console.log('finished training.');

  state = 'prediction';

}

​

​

function mousePressed() {

​

  let inputs = {

    x: mouseX,

    y: mouseY

  }

​

  if (state == 'collection') {

    let target = {

      label: targetLabel

    }

    model.addData(inputs, target);

    stroke(0);

    noFill();

    ellipse(mouseX, mouseY, 24);

    fill(0);

    noStroke();

    textAlign(CENTER, CENTER);

    text(targetLabel, mouseX, mouseY);

​

    wave.freq(notes[targetLabel]);

    env.play();

​

  } else if (state == 'prediction') {

    model.classify(inputs, gotResults);

​

  }

​

}

​

function gotResults(error, results) {

  if (error) {

    console.error(error);

    return;

  }

  console.log(results);

  stroke(0);

  fill(0, 0, 255, 100);

  ellipse(mouseX, mouseY, 24);

  fill(0);

  noStroke();

  textAlign(CENTER, CENTER);

  let label = results[0].label;

  text(label, mouseX, mouseY);

  wave.freq(notes[label]);

  env.play();

}