// NYUSH F24 - gohai

​

let imageSize = 64; // the CNN will be created for 64x64 pixels

let imageChannels = 4; // and four color channels per pixel

let nn;

​

function setup() {

  createCanvas(400, 400);

  background(255);

​

  // For this example to work across all browsers

  // "webgl" or "cpu" needs to be set as the backend

  ml5.setBackend("webgl");

​

  let options = {

    task: "imageClassification", // necessary

    inputs: [imageSize, imageSize, imageChannels], // necessary

    debug: true,

  };

​

  nn = ml5.neuralNetwork(options);

}

​

function draw() {

  // simple drawing tool to demo

  if (mouseIsPressed) {

    fill(0);

    noStroke();

    ellipse(mouseX, mouseY, 50, 50);

  }

}

​

function keyPressed() {

  // resize the canvas to 64x64 pixels

  let resizedImage = get();

  resizedImage.resize(imageSize, imageSize);

​

  if (key == "c") {

    // classify

    nn.classify({ image: resizedImage }, doneClassifying);

  } else if (key == "e") {

    // erase

    background(255);

  } else if (key == "t") {

    // train

    let options = {

      epochs: 64,

    };

    nn.normalizeData();

    nn.train(options, doneTraining);

  } else {

    // add label

    nn.addData({ image: resizedImage }, [key]);

    console.log("Added image for class " + key);

    background(255);

  }

}

​

function doneTraining() {

  console.log("Done training!");

}

​

function doneClassifying(results) {

  console.log(results[0].label + " @ " + results[0].confidence);

}

​