IMU classifier 

works with any webUSB compatible Arduino board and IMu sensor sending accell and gyro outputs. 

​

This example uses the LSM9DS1 and requires this Arduino code uplaoded to work -->

​

https://github.com/j3nsykes/creativeML2020/tree/master/PhysicalTeachableMachines/ML5_neuralNet/Arduino/IMU_visualiser

​

wait 5 seconds at beginning for serial to kick in. 

once you see the data streaming in you can start recording smaples. 

​

Neural Net structure built from Dan Shiffman example on the Coding Train

https://thecodingtrain.com/learning/ml5/6.1-train-your-own.html

*/

​

var sensorData = 0;

var val;

var data = 0; //give initial values. 

var totalInputs = 3; //amount of sensors input X,Y,Z

​

//ML variables

let confidence;

let classificationResult = "waiting";

let brain;

let _x = 0

let _y = 0

let _z = 0;

let sample = 0;

let training = false;

let lastTraining = false;

let recording = false;

let font;

let  inputs=[];

let bg='#FFDE03';

​

function preload() {

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

​

}

​

function setup() {

  let canvas = createCanvas(400, 400);

  canvas.parent('sketch-holder');

​

  samplesP = createP('waiting for samples');

  samplesP.position(130, 85);

  // Create ethe model

  const options = {

    inputs: ['x', 'y', 'z'], // add each electron

    outputs: 3, // add amount of output labels

    debug: true,

    task: 'classification'

  }

  brain = ml5.neuralNetwork(options);

​

  textFont(font);

  background(bg);

}

​

// Train the model

function trainModel() {

  console.log('train model');

  brain.normalizeData();

​

  let options = {

    epochs: 25

  }

​

  brain.train(options, finishedTraining);

}

​

// When the model is trained

function finishedTraining() {

  classify();

}

​

// Classify

function classify() {

  //console.log('classify');

​

  let inputs = getInputs();

  brain.classify(inputs, gotResults);

​

}

​

// Got a result

function gotResults(error, results) {

  if (error) {

    console.error(error);

    return;

  }

​

  classificationResult = results[0].label;

  confidence = results[0].confidence;

  updateClassification(results); //filter results

​

  // Classify again

  classify();

}

​

function updateClassification(results) {

​

  //filter the results

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

    return e.label == 'LINE';

  });

  //console.log(idA[0].confidence);

  labelA = idA[0].label;

  confA = idA[0].confidence;

​

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

    return e.label == 'TRIANGLE';

  });

  //console.log(idB[0].confidence);

  labelB = idB[0].label;

  confB = idB[0].confidence;

​

  //comment in and out for less/more options

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

    return e.label == 'CIRCLE';

  });

  // console.log(idC[0].confidence);

  labelC = idC[0].label;

  confC = idC[0].confidence;

​

  /*

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

      return e.label == 'D';

    });

​

    // console.log(idD[0].confidence);

    labelD = idD[0].label;

    confD = idD[0].confidence;

  */

}

​

function getInputs() {

  //console.log(sensorData);

  sensorData = str(sensorData); //parse to string. 

  let splitString = splitTokens(sensorData, ';'); //seperate at the ';'

​

  if (millis() >= 5000) { //allow for Serial to start stream at beginning. 

    //  console.log(splitString);

    data = JSON.parse(splitString); //parse to JSON object 

    inputs = [];

    fill(0);

    textSize(22);

    text('data.tX: ' + data.tX, 50, 50);

    text('data.tY: ' + data.tY, 50, 100);

    text('data.tZ: ' + data.tZ, 50, 150);

    _x = data.tX;

    _y = data.tY;

    _z = data.tZ;

    inputs.push(_x);

    inputs.push(_y);

    inputs.push(_y);

  }

  return inputs;

}

​

// Add a data record

function addExample() {

​

  samplesP.html('samples: ' + sample++); //see amount of samples collected. 

  let inputs = getInputs();

  //console.log(inputs)

  // Get frequency

  let target = classifierLabel;

  // Add data

  brain.addData(inputs, [target]);

​

}

​

function draw() {

  background(bg);

  getInputs();

  controlsCheck();

​

  fill(255);

  textSize(18);

  text(classificationResult, 50, 300);

​

  ////////////////// ADD YOUR INTTERACTION !!!!! ///////////////////////////

  //check classification results

//   if (classificationResult === "A") {

//     fill(255, 0, 0, 100);

//     ellipse(75, 350, 50, 50);

//   }

​

//   if (classificationResult === "B") {

//     fill(0, 255, 0, 100);

//     ellipse(75, 350, 50, 50);

//   }

​

​

​

}

​

​

function controlsCheck() {

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

    if (training) {

      //console.log('training');

      trainModel();

    }

  }

  lastTraining = training;

​

​

  if (recording) {

    addExample(); //add samples if button checked 

    // console.log('recording');

  }

}