今天在GitHub上发现了一个好玩的代码，短短几百行代码就实现了从复杂的背景视频中去除人物，不得不说这位大佬比较厉害。
这个项目只需要在网络浏览器中使用JavaScript，用200多行TensorFlow.js代码，就可以实时让视频画面中的人物对象从复杂的背景中凭空消失！
耐不住激动就赶快试了一下，虽然没有官方提供的那么完美，但已经很不错了
项目的GitHub地址：https://github.com/jasonmayes/Real-Time-Person-Removal
（一）效果演示
看下官方的演示视频：
我测试的：
（二）代码
GitHub网站最近访问比较慢，这里我直接放上代码
index.html
<!DOCTYPE html>
<html lang="en">
<head>
<title>Disappearing People Project</title>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="author" content="Jason Mayes">
<!-- Import the webpage's stylesheet -->
<link rel="stylesheet" href="style.css">
<!-- Import TensorFlow.js library -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow/tfjs/dist/tf.min.js" type="text/javascript"></script>
</head>
<body>
<h1>Disappearing People Project</h1>
<header class="note">
<h2>Removing people from complex backgrounds in real time using TensorFlow.js</h2>
</header>
<h2>How to use</h2>
<p>Please wait for the model to load before trying the demos below at which point they will become visible when ready to use.</p>
<p>Here is a video of what you can expect to achieve using my custom algorithm. The top is the actual footage, the bottom video is with the real time removal of people working in JavaScript!</p>
<iframe width="540" height="812" src="https://www.youtube.com/embed/0LqEuc32uTc?controls=0&autoplay=1" frameborder="0" allow="accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>
<section id="demos" class="invisible">
<h2>Demo: Webcam live removal</h2>
<p>Try this out using your webcam. Stand a few feet away from your webcam and start walking around... Watch as you slowly disappear in the bottom preview.</p>
<div id="liveView" class="webcam">
<button id="webcamButton">Enable Webcam</button>
<video id="webcam" autoplay></video>
</div>
</section>
<!-- Include the Glitch button to show what the webpage is about and
to make it easier for folks to view source and remix -->
<div class="glitchButton" style="position:fixed;top:20px;right:20px;"></div>
<script src="https://button.glitch.me/button.js"></script>
<!-- Load the bodypix model to recognize body parts in images -->
<script src="https://cdn.jsdelivr.net/npm/@tensorflow-models/[email protected]"></script>
<!-- Import the page's JavaScript to do some stuff -->
<script src="script.js" defer></script>
</body>
</html>
stript.js
/**
* @license
* Copyright 2018 Google LLC. All Rights Reserved.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
* =============================================================================
*/
/********************************************************************
* Real-Time-Person-Removal Created by Jason Mayes 2020.
*
* Get latest code on my Github:
* https://github.com/jasonmayes/Real-Time-Person-Removal
*
* Got questions? Reach out to me on social:
* Twitter: @jason_mayes
* LinkedIn: https://www.linkedin.com/in/creativetech
********************************************************************/
const video = document.getElementById('webcam');
const liveView = document.getElementById('liveView');
const demosSection = document.getElementById('demos');
const DEBUG = false;
// An object to configure parameters to set for the bodypix model.
// See github docs for explanations.
const bodyPixProperties = {
architecture: 'MobileNetV1',
outputStride: 16,
multiplier: 0.75,
quantBytes: 4
};
// An object to configure parameters for detection. I have raised
// the segmentation threshold to 90% confidence to reduce the
// number of false positives.
const segmentationProperties = {
flipHorizontal: false,
internalResolution: 'high',
segmentationThreshold: 0.9
};
// Must be even. The size of square we wish to search for body parts.
// This is the smallest area that will render/not render depending on
// if a body part is found in that square.
const SEARCH_RADIUS = 300;
const SEARCH_OFFSET = SEARCH_RADIUS / 2;
// RESOLUTION_MIN should be smaller than SEARCH RADIUS. About 10x smaller seems to
// work well. Effects overlap in search space to clean up body overspill for things
// that were not classified as body but infact were.
const RESOLUTION_MIN = 20;
// Render returned segmentation data to a given canvas context.
function processSegmentation(canvas, segmentation) {
var ctx = canvas.getContext('2d');
// Get data from our overlay canvas which is attempting to estimate background.
var imageData = ctx.getImageData(0, 0, canvas.width, canvas.height);
var data = imageData.data;
// Get data from the live webcam view which has all data.
var liveData = videoRenderCanvasCtx.getImageData(0, 0, canvas.width, canvas.height);
var dataL = liveData.data;
// Now loop through and see if pixels contain human parts. If not, update
// backgound understanding with new data.
for (let x = RESOLUTION_MIN; x < canvas.width; x += RESOLUTION_MIN) {
for (let y = RESOLUTION_MIN; y < canvas.height; y += RESOLUTION_MIN) {
// Convert xy co-ords to array offset.
let n = y * canvas.width + x;
let foundBodyPartNearby = false;
// Let's check around a given pixel if any other pixels were body like.
let yMin = y - SEARCH_OFFSET;
yMin = yMin < 0 ? 0: yMin;
let yMax = y + SEARCH_OFFSET;
yMax = yMax > canvas.height ? canvas.height : yMax;
let xMin = x - SEARCH_OFFSET;
xMin = xMin < 0 ? 0: xMin;
let xMax = x + SEARCH_OFFSET;
xMax = xMax > canvas.width ? canvas.width : xMax;
for (let i = xMin; i < xMax; i++) {
for (let j = yMin; j < yMax; j++) {
let offset = j * canvas.width + i;
// If any of the pixels in the square we are analysing has a body
// part, mark as contaminated.
if (segmentation.data[offset] !== 0) {
foundBodyPartNearby = true;
break;
}
}
}
// Update patch if patch was clean.
if (!foundBodyPartNearby) {
for (let i = xMin; i < xMax; i++) {
for (let j = yMin; j < yMax; j++) {
// Convert xy co-ords to array offset.
let offset = j * canvas.width + i;
data[offset * 4] = dataL[offset * 4];
data[offset * 4 + 1] = dataL[offset * 4 + 1];
data[offset * 4 + 2] = dataL[offset * 4 + 2];
data[offset * 4 + 3] = 255;
}
}
} else {
if (DEBUG) {
for (let i = xMin; i < xMax; i++) {
for (let j = yMin; j < yMax; j++) {
// Convert xy co-ords to array offset.
let offset = j * canvas.width + i;
data[offset * 4] = 255;
data[offset * 4 + 1] = 0;
data[offset * 4 + 2] = 0;
data[offset * 4 + 3] = 255;
}
}
}
}
}
}
ctx.putImageData(imageData, 0, 0);
}
// Let's load the model with our parameters defined above.
// Before we can use bodypix class we must wait for it to finish
// loading. Machine Learning models can be large and take a moment to
// get everything needed to run.
var modelHasLoaded = false;
var model = undefined;
model = bodyPix.load(bodyPixProperties).then(function (loadedModel) {
model = loadedModel;
modelHasLoaded = true;
// Show demo section now model is ready to use.
demosSection.classList.remove('invisible');
});
/********************************************************************
// Continuously grab image from webcam stream and classify it.
********************************************************************/
var previousSegmentationComplete = true;
// Check if webcam access is supported.
function hasGetUserMedia() {
return !!(navigator.mediaDevices &&
navigator.mediaDevices.getUserMedia);
}
// This function will repeatidly call itself when the browser is ready to process
// the next frame from webcam.
function predictWebcam() {
if (previousSegmentationComplete) {
// Copy the video frame from webcam to a tempory canvas in memory only (not in the DOM).
videoRenderCanvasCtx.drawImage(video, 0, 0);
previousSegmentationComplete = false;
// Now classify the canvas image we have available.
model.segmentPerson(videoRenderCanvas, segmentationProperties).then(function(segmentation) {
processSegmentation(webcamCanvas, segmentation);
previousSegmentationComplete = true;
});
}
// Call this function again to keep predicting when the browser is ready.
window.requestAnimationFrame(predictWebcam);
}
// Enable the live webcam view and start classification.
function enableCam(event) {
if (!modelHasLoaded) {
return;
}
// Hide the button.
event.target.classList.add('removed');
// getUsermedia parameters.
const constraints = {
video: true
};
// Activate the webcam stream.
navigator.mediaDevices.getUserMedia(constraints).then(function(stream) {
video.addEventListener('loadedmetadata', function() {
// Update widths and heights once video is successfully played otherwise
// it will have width and height of zero initially causing classification
// to fail.
webcamCanvas.width = video.videoWidth;
webcamCanvas.height = video.videoHeight;
videoRenderCanvas.width = video.videoWidth;
videoRenderCanvas.height = video.videoHeight;
let webcamCanvasCtx = webcamCanvas.getContext('2d');
webcamCanvasCtx.drawImage(video, 0, 0);
});
video.srcObject = stream;
video.addEventListener('loadeddata', predictWebcam);
});
}
// We will create a tempory canvas to render to store frames from
// the web cam stream for classification.
var videoRenderCanvas = document.createElement('canvas');
var videoRenderCanvasCtx = videoRenderCanvas.getContext('2d');
// Lets create a canvas to render our findings to the DOM.
var webcamCanvas = document.createElement('canvas');
webcamCanvas.setAttribute('class', 'overlay');
liveView.appendChild(webcamCanvas);
// If webcam supported, add event listener to button for when user
// wants to activate it.
if (hasGetUserMedia()) {
const enableWebcamButton = document.getElementById('webcamButton');
enableWebcamButton.addEventListener('click', enableCam);
} else {
console.warn('getUserMedia() is not supported by your browser');
}
style.js
/**
* @license
* Copyright 2018 Google LLC. All Rights Reserved.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
* =============================================================================
*/
/******************************************************
* Stylesheet by Jason Mayes 2020.
*
* Get latest code on my Github:
* https://github.com/jasonmayes/Real-Time-Person-Removal
* Got questions? Reach out to me on social:
* Twitter: @jason_mayes
* LinkedIn: https://www.linkedin.com/in/creativetech
*****************************************************/
body {
font-family: helvetica, arial, sans-serif;
margin: 2em;
color: #3D3D3D;
}
h1 {
font-style: italic;
color: #FF6F00;
}
h2 {
clear: both;
}
em {
font-weight: bold;
}
video {
clear: both;
display: block;
}
section {
opacity: 1;
transition: opacity 500ms ease-in-out;
}
header, footer {
clear: both;
}
button {
z-index: 1000;
position: relative;
}
.removed {
display: none;
}
.invisible {
opacity: 0.2;
}
.note {
font-style: italic;
font-size: 130%;
}
.webcam {
position: relative;
}
.webcam, .classifyOnClick {
position: relative;
float: left;
width: 48%;
margin: 2% 1%;
cursor: pointer;
}
.webcam p, .classifyOnClick p {
position: absolute;
padding: 5px;
background-color: rgba(255, 111, 0, 0.85);
color: #FFF;
border: 1px dashed rgba(255, 255, 255, 0.7);
z-index: 2;
font-size: 12px;
}
.highlighter {
background: rgba(0, 255, 0, 0.25);
border: 1px dashed #fff;
z-index: 1;
position: absolute;
}
.classifyOnClick {
z-index: 0;
position: relative;
}
.classifyOnClick canvas, .webcam canvas.overlay {
opacity: 1;
top: 0;
left: 0;
z-index: 2;
}
#liveView {
transform-origin: top left;
transform: scale(1);
}
