Created
March 25, 2018 11:28
-
-
Save mac2000/fc54e6d6bdcbfde28b03dc2a43611270 to your computer and use it in GitHub Desktop.
brain.js demo
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
<!DOCTYPE html> | |
<html> | |
<head> | |
<title>brain.js</title> | |
<script src="https://cdn.rawgit.com/BrainJS/brain.js/develop/browser.js"></script> | |
<script> | |
function DrawableCanvas(el) { | |
const px = 10 | |
const ctx = el.getContext('2d') | |
let x = [] | |
let y = [] | |
let moves = [] | |
let isPainting = false | |
const clear = () => ctx.clearRect(0, 0, ctx.canvas.width, ctx.canvas.height) | |
const addPoint = (_x, _y, isMoving) => { | |
x.push(_x) | |
y.push(_y) | |
moves.push(isMoving) | |
} | |
const redraw = () => { | |
clear() | |
ctx.strokeStyle = 'red' | |
ctx.lineJoin = 'round' | |
ctx.lineWidth = px | |
for (let i = 0; i < moves.length; i++) { | |
ctx.beginPath() | |
if (moves[i] && i) { | |
ctx.moveTo(x[i - 1], y[i - 1]) | |
} else { | |
ctx.moveTo(x[i] - 1, y[i]) | |
} | |
ctx.lineTo(x[i], y[i]) | |
ctx.closePath() | |
ctx.stroke() | |
} | |
} | |
const drawLine = (x1, y1, x2, y2, color = 'lightgray') => { | |
ctx.beginPath() | |
ctx.strokeStyle = color | |
ctx.lineJoin = 'miter' | |
ctx.lineWidth = 1 | |
ctx.moveTo(x1, y1) | |
ctx.lineTo(x2, y2) | |
ctx.stroke() | |
} | |
const grid = () => { | |
const w = el.clientWidth | |
const h = el.clientHeight | |
const p = el.clientWidth / px | |
const xStep = w / p | |
const yStep = h / p | |
for(let x = 0; x < w; x += xStep) { | |
drawLine(x, 0, x, h) | |
} | |
for(let y = 0; y < h; y += yStep) { | |
drawLine(0, y, w, y) | |
} | |
} | |
const cell = (x, y, w, h) => { | |
ctx.fillStyle = 'blue' | |
ctx.strokeStyle = 'blue' | |
ctx.lineJoin = 'miter' | |
ctx.lineWidth = 1 | |
ctx.rect(x, y, w, h) | |
ctx.fill() | |
} | |
this.reset = () => { | |
isPainting = false | |
x = [] | |
y = [] | |
moves = [] | |
clear() | |
} | |
this.getVector = (debug = false) => { | |
const w = el.clientWidth | |
const h = el.clientHeight | |
const p = el.clientWidth / px | |
const xStep = w / p | |
const yStep = h / p | |
const vector = [] | |
for(let x = 0; x < w; x += xStep) { | |
for(let y = 0; y < h; y += yStep) { | |
const data = ctx.getImageData(x, y, xStep, yStep) | |
let nonEmptyPixelsCount = 0 | |
for(let i = 0; i < data.data.length; i += 4) { | |
const isEmpty = data.data[i] === 0 | |
if (!isEmpty) { | |
nonEmptyPixelsCount += 1 | |
} | |
} | |
if (nonEmptyPixelsCount > 1 && debug) { | |
cell(x, y, xStep, yStep) | |
} | |
vector.push(nonEmptyPixelsCount > 1 ? 1 : 0) | |
} | |
} | |
if (debug) { | |
grid() | |
} | |
return vector | |
} | |
el.addEventListener('mousedown', event => { | |
const bounds = event.target.getBoundingClientRect() | |
const x = event.clientX - bounds.left | |
const y = event.clientY - bounds.top | |
isPainting = true | |
addPoint(x, y, false) | |
redraw() | |
}) | |
el.addEventListener('mousemove', event => { | |
const bounds = event.target.getBoundingClientRect() | |
const x = event.clientX - bounds.left | |
const y = event.clientY - bounds.top | |
if (isPainting) { | |
addPoint(x, y, true) | |
redraw() | |
} | |
}) | |
el.addEventListener('mouseup', () => { | |
isPainting = false | |
}) | |
el.addEventListener('mouseleave', () => { | |
isPainting = false | |
}) | |
} | |
</script> | |
</head> | |
<body> | |
<table align="center" cellspacing="0" cellpadding="1" border="0"> | |
<caption> | |
<h3>Step 1: Prepare data</h3> | |
<p>Draw 3 positive and 3 negative images</p> | |
</caption> | |
<tr> | |
<th>Positive</th> | |
<td> | |
<canvas id="p1" width="200" height="200" style="border: 1px solid black; cursor: default; display: block;"></canvas> | |
</td> | |
<td> | |
<canvas id="p2" width="200" height="200" style="border: 1px solid black; cursor: default; display: block;"></canvas> | |
</td> | |
<td> | |
<canvas id="p3" width="200" height="200" style="border: 1px solid black; cursor: default; display: block;"></canvas> | |
</td> | |
</tr> | |
<tr> | |
<th>Negative</th> | |
<td> | |
<canvas id="n1" width="200" height="200" style="border: 1px solid black; cursor: default; display: block;"></canvas> | |
</td> | |
<td> | |
<canvas id="n2" width="200" height="200" style="border: 1px solid black; cursor: default; display: block;"></canvas> | |
</td> | |
<td> | |
<canvas id="n3" width="200" height="200" style="border: 1px solid black; cursor: default; display: block;"></canvas> | |
</td> | |
</tr> | |
</table> | |
<table width="100%" align="center"> | |
<caption> | |
<h3>Step 2: Train Model</h3> | |
<button id="train">Train</button> | |
<label><input type="checkbox" id="dbg"/> debug</label> | |
</caption> | |
<tr> | |
<td align="center"> | |
<table> | |
<tbody id="res" style="display: none"> | |
<tr> | |
<th>Error</th> | |
<td id="err"></td> | |
</tr> | |
<tr> | |
<th>Iterations</th> | |
<td id="iterations"></td> | |
</tr> | |
</tbody> | |
</table> | |
</td> | |
</tr> | |
</table> | |
<table align="center"> | |
<caption> | |
<h3>Step 3: Evaluate Model</h3> | |
<button id="guess">Guess</button> | |
</caption> | |
<tr> | |
<td align="center"> | |
<canvas id="g" width="200" height="200" style="border: 1px solid black; cursor: default; display: block;"></canvas> | |
</td> | |
</tr> | |
</table> | |
<script> | |
const pc1 = new DrawableCanvas(document.getElementById('p1')) | |
const pc2 = new DrawableCanvas(document.getElementById('p2')) | |
const pc3 = new DrawableCanvas(document.getElementById('p3')) | |
const nc1 = new DrawableCanvas(document.getElementById('n1')) | |
const nc2 = new DrawableCanvas(document.getElementById('n2')) | |
const nc3 = new DrawableCanvas(document.getElementById('n3')) | |
const gc = new DrawableCanvas(document.getElementById('g')) | |
const net = new brain.NeuralNetwork() | |
train.addEventListener('click', () => { | |
const data = [] | |
data.push({ input: pc1.getVector(dbg.checked), output: {positive: 1} }) | |
data.push({ input: pc2.getVector(dbg.checked), output: {positive: 1} }) | |
data.push({ input: pc3.getVector(dbg.checked), output: {positive: 1} }) | |
data.push({ input: nc1.getVector(dbg.checked), output: {negative: 1} }) | |
data.push({ input: nc2.getVector(dbg.checked), output: {negative: 1} }) | |
data.push({ input: nc3.getVector(dbg.checked), output: {negative: 1} }) | |
const result = net.train(data, {log: true}) | |
err.innerHTML = result.error | |
iterations.innerHTML = result.iterations | |
res.removeAttribute('style') | |
}) | |
guess.addEventListener('click', () => { | |
const result = brain.likely(gc.getVector(), net) | |
alert(result) | |
gc.reset() | |
}) | |
</script> | |
</body> | |
</html> |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
https://output.jsbin.com/mofaduk, example link