// Cloned by Abdelshafa Abdala on 17 Nov 2021 from World "Character recognition neural network (clone by Fergus)" by Fergus
// Please leave this clone trail here.
const PIXELS=28,
PIXELSSQUARED=PIXELS*PIXELS,
NOTRAIN=6e4,
NOTEST=1e4,
noinput=PIXELSSQUARED,
nohidden=256,
nooutput=10,
learningrate=.1;
let learning_array=[],
learning_target=5,
do_training=!0;
const TRAINPERSTEP=24,
TESTPERSTEP=4,
ZOOMFACTOR=6,
ZOOMPIXELS=6*PIXELS;
let percent=.1;
const canvaswidth=PIXELS+ZOOMPIXELS+50,
canvasheight=3*ZOOMPIXELS+100,
DOODLE_THICK=14,
DOODLE_BLUR=.2;
let mnist,nn,doodle,demo,
trainrun=1,
train_index=0,
testrun=1,
test_index=0,
total_tests=0,
total_correct=0,
doodle_exists=!1,
demo_exists=!1,
mousedrag=!1;
var train_inputs,test_inputs,demo_inputs,doodle_inputs,doodle_inputs_c,
thehtml;
function randomWeight()
{
return AB.randomFloatAtoB(-.5,.5)}$("#runheaderbox").css({"max-height":"95vh"}),
thehtml="<hr> <h1> 1. Doodle </h1> Top row: Doodle (left) and shrunk (right). <br> Draw your doodle in top LHS. <button onclick='wipeDoodle();' class='normbutton' >Clear doodle</button> <br> ",
AB.msg(thehtml,1),thehtml="<hr> <h1> 2. Training </h1> Middle row: Training image magnified (left) and original (right). <br> <button onclick='do_training = false;' class='normbutton' >Stop training</button> <br> ",
AB.msg(thehtml,3),thehtml="<h3> Hidden tests </h3> ",
AB.msg(thehtml,5),thehtml="<hr> <h1> 3. Demo </h1> Bottom row: Test image magnified (left) and original (right). <br> The network is <i>not</i> trained on any of these images. <br> <button onclick='makeDemo();' class='normbutton' >Demo test image</button> <br> ",
AB.msg(thehtml,7);
const greenspan="<span style='font-weight:bold; font-size:x-large; color:darkgreen'> ";
function setup(){createCanvas(canvaswidth,canvasheight),(doodle=createGraphics(ZOOMPIXELS,ZOOMPIXELS)).pixelDensity(.15),
AB.loadingScreen(),$.getScript("/uploads/codingtrain/matrix.js",
function(){$.getScript("/uploads/fergus24/nn.js",
function(){$.getScript("/uploads/codingtrain/mnist.js",
function(){console.log("All JS loaded"),(nn=new NeuralNetwork(noinput,nohidden,nooutput)).setLearningRate(learningrate),loadData()})})})}
function loadData(){loadMNIST(
function(t){mnist=t,
console.log("All data loaded into mnist object:"),
console.log(mnist),AB.removeLoading()})}
function getImage(t){
let e=createImage(PIXELS,PIXELS);e.loadPixels();
for(let n=0;n<PIXELSSQUARED;n++){
let o=1*t[n],i=4*n;e.pixels[i+0]=o,e.pixels[i+1]=o,e.pixels[i+2]=o,e.pixels[i+3]=255}
return e.updatePixels(),e}
function getInputs(t){let e=[];
for(let n=0;n<PIXELSSQUARED;n++){let o=t[n];e[n]=o/255}return e}
function trainit(t){let e=mnist.train_images[train_index],n=mnist.train_labels[train_index];
if(t){
var o=getImage(e);image(o,0,
ZOOMPIXELS+ZOOMPIXELS/5,
ZOOMPIXELS,ZOOMPIXELS),
image(o,ZOOMPIXELS+50,
ZOOMPIXELS+ZOOMPIXELS/5,
PIXELS,PIXELS)}
let i=getInputs(e),r=[0,0,0,0,0,0,0,0,0,0];r[n]=1,train_inputs=i,nn.train(i,r),
thehtml=" trainrun: "+trainrun+"<br> no: "+train_index,AB.msg(thehtml,4),++train_index==NOTRAIN&&(train_index=0,
console.log("finished trainrun: "+trainrun),trainrun++)}
function testit(){
let t=mnist.test_images[test_index],e=mnist.test_labels[test_index],n=getInputs(t);test_inputs=n;
let o=findMax(nn.predict(n));total_tests++,o==e&&total_correct++,percent=total_correct/total_tests*100;
let i=Math.log(percent/100)/Math.log(60)*-2;i>.5&&(i=.5),
console.log("modified learning rate "+i/1),nn.setLearningRate(i/1),percent>learning_target&&(learning_array[learning_target/5]=total_tests,
console.log("learning target"+learning_target),
console.log("learning array"+learning_array),learning_target+=5),
thehtml=" testrun: "+testrun+"<br> no: "+total_tests+" <br> correct: "+total_correct+"<br> score: "+greenspan+percent.toFixed(2)+"</span>",
AB.msg(thehtml,6),++test_index==NOTEST&&(
console.log("finished testrun: "+testrun+" score: "+percent.toFixed(2)),testrun++,test_index=0,total_tests=0,total_correct=0)}
function find12(t){let e=0,n=0,o=0,i=0;for(let r=0;r<t.length;r++)t[r]>o?(e=r,o=t[r]):t[r]>i&&(n=r,i=t[r]);
return[e,n]}
function findMax(t){
let e=0,n=0;for(
let o=0;o<t.length;o++)t[o]>n&&(e=o,n=t[o]);
return e}
function draw(){
if(void 0!==mnist){
if(background("#ffffcc"),
bg=createGraphics(1.1*ZOOMPIXELS,1.1*ZOOMPIXELS),bg.background("red"),
image(bg,0,0,1.05*ZOOMPIXELS,1.05*ZOOMPIXELS),
image(bg,ZOOMPIXELS+50,0,1.05*PIXELS,1.05*PIXELS),
image(bg,0,1*ZOOMPIXELS+1*ZOOMPIXELS/5,1.05*ZOOMPIXELS,1.05*ZOOMPIXELS),
image(bg,ZOOMPIXELS+50,1*ZOOMPIXELS+1*ZOOMPIXELS/5,1.05*ZOOMPIXELS,1.05*PIXELS),
image(bg,0,canvasheight-1.05*ZOOMPIXELS,1.05*ZOOMPIXELS,1.05*ZOOMPIXELS),
image(bg,ZOOMPIXELS+50,canvasheight-1.05*ZOOMPIXELS,1.05*PIXELS,1.05*PIXELS),
do_training){for(let t=0;t<TRAINPERSTEP;t++)trainit(0===t);
for(let t=0;t<TESTPERSTEP;t++)testit()}
if(demo_exists&&(drawDemo(),guessDemo()),doodle_exists&&(drawDoodle(),guessDoodle()),mouseIsPressed){
var t=ZOOMPIXELS+0;
mouseX<t&&mouseY<t&&pmouseX<t&&pmouseY<t&&(mousedrag=!
0,doodle_exists=!0,
doodle.stroke("white"),doodle.strokeWeight(DOODLE_THICK),doodle.line(mouseX,mouseY,pmouseX,pmouseY))}
else mousedrag&&(mousedrag=!1,
doodle.filter(BLUR,DOODLE_BLUR))}}
function makeDemo(){demo_exists=!0;
var t=AB.randomIntAtoB(0,NOTEST-1);demo=mnist.test_images[t];
var e=mnist.test_labels[t];thehtml="Test image no: "+t+"<br>Classification: "+e+"<br>",AB.msg(thehtml,8)}
function drawDemo(){
var t=getImage(demo);image(t,0,
canvasheight-ZOOMPIXELS+ZOOMPIXELS/100,
ZOOMPIXELS,ZOOMPIXELS),image(t,ZOOMPIXELS+50,
canvasheight-ZOOMPIXELS+ZOOMPIXELS/100,
PIXELS,PIXELS)}
function guessDemo()
{
let t=getInputs(demo);demo_inputs=t;
let e=findMax(nn.predict(t));thehtml=" We classify it as: "+greenspan+e+"</span>",AB.msg(thehtml,9)}
function drawDoodle(){
let t=doodle.get();image(t,0,0,ZOOMPIXELS,ZOOMPIXELS),image(t,ZOOMPIXELS+50,0,PIXELS,PIXELS)}
function guessDoodle(){let t=doodle.get();t.resize(PIXELS,PIXELS),t.loadPixels();
let e=[];
for(let n=0;n<PIXELSSQUARED;n++)e[n]=t.pixels[4*n]/255;
let n=0,o=0,i=0,r=0,a=e;
if(!1===mousedrag){a=Array.apply(null,Array(PIXELS*PIXELS)).map(Number.prototype.valueOf,0);
for(let t=0;t<PIXELS;t++)for(let o=0;o<PIXELS;o++)e[t*PIXELS+o]>0&&0===n&&(n=t);
for(h=PIXELS-1;h>0;h--)
for(w=0;w<PIXELS;w++)e[h*PIXELS+w]>0&&0===o&&(o=h);
for(w=0;w<PIXELS;w++)
for(h=0;h<PIXELS;h++)e[h*PIXELS+w]>0&&0===i&&(i=w);
for(w=PIXELS-1;w>0;w--)for(h=0;h<PIXELS;h++)e[h*PIXELS+w]>0&&0===r&&(r=w);
let t=Math.round(PIXELS/2-(o-n)/2-n),s=Math.round(PIXELS/2-(r-i)/2-i);
for(let n=0;n<PIXELS;n++)
for(let o=0;o<PIXELS;o++)e[n*PIXELS+o]>0&&(a[(n+t)*PIXELS+(o+s)]=e[n*PIXELS+o])}doodle_inputs=e,doodle_inputs_c=a;
let s=find12(nn.predict(a));thehtml=" We classify it as: "+greenspan+s[0]+"</span> <br> No.2 guess is: "+greenspan+s[1]+"</span>",AB.msg(thehtml,2)}
function wipeDoodle(){doodle_exists=!1,doodle.background("black")}
function showInputs(t){var e="";for(let n=0;n<t.length;n++){n%PIXELS==0&&(e+="\n"),e=e+" "+t[n].toFixed(2)}
console.log(e)}