//

// This software is released under the MIT License.

// https://opensource.org/licenses/MIT

​

/* ===

ml5 Example

Image classification using MobileNet and p5.js

This example uses a callback pattern to create the classifier

=== */

let nn;

​

function setup() {

​

  const options = {

    task: 'imageClassification',

    inputs: [2, 2, 4],

    debug: true

  }

  nn = ml5.neuralNetwork(options);

​

  addData();

  nn.train({

    epochs: 20,

    batchSize: 2

  }, finishedTraining)

​

}

​

function finishedTraining() {

  nn.classify([

    0, 0, 255, 255, 0, 0, 255, 255,

    0, 0, 255, 255, 0, 0, 255, 255

  ], gotResults)

​

}

​

function gotResults(err, result) {

  if (err) {

    console.error(err)

  }

  console.log(result)

}

​

function addData() {

  const myData = [{

      label: "red-square",

      value: [

        255, 0, 0, 255, 255, 0, 0, 255,

        255, 0, 0, 255, 255, 0, 0, 255

      ]

    },

    {

      label: "green-square",

      value: [

        0, 255, 0, 255, 0, 255, 0, 255,

        0, 255, 0, 255, 0, 255, 0, 255

      ]

    },

    {

      label: "blue-square",

      value: [

        0, 0, 255, 255, 0, 0, 255, 255,

        0, 0, 255, 255, 0, 0, 255, 255

      ]

    }

​

  ]

​

  // method 1: adding data as objects

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

    const item = myData[i];

    const xInputObj = {

      pixelArray: item.value

    }

​

    const yInputObj = {

      label: item.label

    }

    nn.addData(xInputObj, yInputObj)

  }

​

  // method 2:adding data as arrays with

  const labelsOptions = {

    inputLabels: ['pixelArray'],

    outputLabels: ['label']

  }

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

    const item = myData[i]; 

    nn.addData([item.value], [item.label], labelsOptions)

  }

}