Skip to content

Instantly share code, notes, and snippets.

@symisc
Last active November 29, 2020 04:37
Show Gist options
  • Save symisc/e633684f0a3f31a0c3272c5fe07a2976 to your computer and use it in GitHub Desktop.
Save symisc/e633684f0a3f31a0c3272c5fe07a2976 to your computer and use it in GitHub Desktop.
Usage example of the WebAssembly Real-Time face detector model - https://pixlab.io/downloads
<!DOCTYPE html>
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<title>WebAssembly Real-Time Face Detection</title>
<meta name="viewport" content="width=device-width, initial-scale=1">
</head>
<style>
body {
font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", "Roboto", "Oxygen", "Ubuntu", "Cantarell", "Fira Sans", "Droid Sans", "Helvetica Neue", sans-serif;
}
</style>
<body>
<div align="center">
<h1>WebAssembly Real-Time Face Detection <small>via <a href="https://sod.pixlab.io" target="_blank">SOD RealNets</a></small></h1>
<button onclick="face_detect()">Start Face Detection</button><hr/>
<canvas id="face-canvas" style="border: 1px solid black;"></canvas>
<video id="face-video" autoplay playsinline hidden="true"></video>
</div>
<!-- This will load the RealNet WebAssembly face model and run its main.
The model must be downloaded from https://pixlab.io/downloads
Once downloaded, just put it on the directory where this HTML
file `usage.html` reside.
For chrome users, you must test the model on actual Web server,
whether served locally (i.e http://127.0.0.1) or remotely.
This is due to the fact that chrome does not allow WebAssembly
modules to be loaded directly from the file system (
Edge and Firefox does not have such issue).
-->
<script src="facemodel.js"></script>
<script>
function face_detect(){
var grayBuf = null; /* Working buffer holding the grayscale conversion of the current video frame */
var recs = null; /* Each detected face coordinates (top, left, width & height) is recorded in this array */
let vidSz = {width: 640, height: 480 , facingMode: "user" };
if (document.body.clientWidth < 600 ){
vidSz = true;
}
navigator.mediaDevices.getUserMedia({
audio: false,
video: vidSz
}).then(function(mediaStream) {
/* Code taken from https://faceio.net JavaScript Widget */
let video = document.getElementById('face-video');
let vidCanvas = document.getElementById('face-canvas');
let vidctx = vidCanvas.getContext('2d');
video.srcObject = mediaStream;
video.onloadedmetadata = function(e) {
vidCanvas.width = video.videoWidth;
vidCanvas.height = video.videoHeight;
video.play();
}
video.addEventListener('play', function () {
let fps = 1000 / 25; /* Render at 25 fps. Feel free to increase to 30 if you are on a PC/Mac or Modern iPhone */
let loop = function() {
if (!video.paused && !video.ended) {
vidctx.drawImage(video, 0, 0);
if( grayBuf === null ){
grayBuf = _realnet_alloc_gray_image_buffer(640, 480);
recs = new Float32Array(Module.HEAPU8.buffer, _realnet_alloc_face_result_array(), _realnet_face_max_detection());
}else{
let imgData = vidctx.getImageData(0,0, vidCanvas.width, vidCanvas.height);
let gray = new Uint8ClampedArray( Module.HEAPU8.buffer, grayBuf, imgData.width * imgData.height);
for (let i = 0; i < imgData.data.length ; i += 4) {
gray[i >> 2] = (imgData.data[i] * 306 + imgData.data[i + 1] * 601 + imgData.data[i + 2] * 117) >> 10;
}
let faces = _realnet_face_detect(grayBuf, imgData.width, imgData.height, 11.0, recs.byteOffset);
/* faces now hold the total number of detected faces and the recs[] array hold their coordinates */
for(let i = 0; i < faces; i += 5 ){
/* Draw a rectangle on each detected face */
vidctx.beginPath();
vidctx.strokeStyle = "#32cd32";
vidctx.strokeRect(recs[i + 0],recs[i + 1],recs[i + 2],recs[i + 3]);
/* recs[i + 4] hold the score (i.e. threshold) of the current face */
console.log("Face_" + i + ": Score: " + recs[i + 4]);
}
}
setTimeout(loop, fps);
}else{
console.log("Video stream stopped");
}
}
/* Start painting on the canvas */
loop();
});
}).catch(function(err){
console.log("Access not granted", err);
});
}
</script>
</body>
</html>
@symisc
Copy link
Author

symisc commented Jan 1, 2019

This will load the RealNet WebAssembly face model and run its main. The model must be downloaded from https://pixlab.io/download. Once downloaded, just put it on the directory where this HTML file usage.html reside.

When you deploy the Webassembly face model on your server, make sure
your HTTP server (Apache, Nginx, etc.) return the appropriate MIME type
for the wasm file extension. Under Apache, simply put the following
directives on your .htaccess or Virtual host configuration:

AddType application/wasm .wasm
AddOutputFilterByType DEFLATE application/wasm

For chrome users, you must test the model on actual Web server,
whether served locally (i.e http://127.0.0.1) or remotely.
This is due to the fact that chrome does not allow WebAssembly
modules to be loaded directly from the file system ( Edge and Firefox does not have such issue).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment