// Cloned by Cian on 5 Dec 2021 from World "Recognise any image (clone by Cian) (clone by Cian)" by Cian
// Please leave this clone trail here.
// Cloned by Cian on 4 Dec 2021 from World "Recognise any image (clone by Cian)" by Cian
// Please leave this clone trail here.
// Cloned by Cian on 4 Dec 2021 from World "Recognise any image" by Starter user
// Please leave this clone trail here.
// 1.) collecting the data (uisng KNN?) 2.) pre-trained model 3). inference/deploy
// saveData() <- data save
// save <- model save
// loadData
// load
// this example uses key-presses, model.saveData(filename, callback)
// ML5 image recognition using MobileNet
// uses AB framework
// enter URL of image at runtime
// no run, pause, step
AB.drawRunControls = false;
let classifier;
let img;
let clearButton;
let canvas;
let doodleClassifier;
// display Pr, pass it into a html elem to display
// asynchronous loads of resources, with callback functions when ready
var classifierLoaded = false;
var imgLoaded = false;
AB.world.newRun = function () {
ABWorld.init('lightblue')
ABWorld.canvas;
//AB.headerRHS();
// uisng splash screen for clear button
// AB.newSplash();
// AB.splashHtml(" <button onclick='clearCanvas();' class=ab-normbutton >Clear</button> <p> " );
// DoodleNet load pretrained
doodleClassifier = ml5.imageClassifier('DoodleNet', modelReady)
AB.msg('model loading')
// allow user inout to be saved
//doodleClassifier.saveData('doodle-notes.json');
function modelReady() {
console.log('model loaded');
// generate res
doodleClassifier.classify(ABWorld.p5canvas, gotResults);
}
// attempting key presses as user interface
function keyPressed(){
if (key == 's') {
console.log('saving doodle')
doodleClassifier.saveData('mouse-notes'); // saving local by default?
}
}
// results----------------------------------------- could use AB.newDiv(id) rather than AB.msg https://ancientbrain.com/docs.ab.php
AB.headerRHS();
AB.msg ( "Running image recognition ... <br> ", 1 );
AB.runReady = true;
function gotResults(error, results){
if (error){
console.log(error);
AB.msg ( "<font color=red> <B> Error recognising doodle. See console for details. </b></font> <br> ", 2 );
return;
}
// display top 2 results
let content = ` ${results[0].label}
${nf(100 * results[0].confidence, 2, 1)}%<br/>
${results[1].label}
${nf(100 * results[1].confidence, 2, 1)}%`;
AB.msg(content);
console.log(content);
doodleClassifier.classify(ABWorld.p5canvas, gotResults);
}
};
// clear func for button
function clearCanvas() {
//ABWorld.canvas('lightblue');
console.log('button pressed')
ABWorld.p5canvas('lightblue');
}
//---- setup -------------------------------------------------------
// Do NOT make a setup function.
// This is done for you in the API. The API setup just creates a canvas.
// Anything else you want to run at the start should go into the following two functions.
function beforesetup() // Optional
{
// Anything you want to run at the start BEFORE the canvas is created
// canvas = createCanvas(400, 400);
// clearButton = createButton('clear');
}
function aftersetup() // Optional
{
// Anything you want to run at the start AFTER the canvas is created
}
//---- draw -------------------------------------------------------
function draw() // Optional
{
// Can put P5 instructions to be executed every step here, or in AB.world.nextStep()
// below line thickness could be affecting classification (overfot to line thickness on train set), originally 8 stroke
if (mouseIsPressed) {
strokeWeight(16);
line(mouseX, mouseY, pmouseX, pmouseY);
}
}