candle-yolo / index.html
lmz's picture
minor UI change (#5)
6b37b03
<html>
<head>
<meta content="text/html;charset=utf-8" http-equiv="Content-Type" />
<title>Candle YOLOv8 Rust/WASM</title>
</head>
<body></body>
</html>
<!doctype html>
<html>
<head>
<meta charset="UTF-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<style>
@import url("https://fonts.googleapis.com/css2?family=Source+Code+Pro:wght@200;300;400&family=Source+Sans+3:wght@100;200;300;400;500;600;700;800;900&display=swap");
html,
body {
font-family: "Source Sans 3", sans-serif;
}
code,
output,
select,
pre {
font-family: "Source Code Pro", monospace;
}
</style>
<script src="https://cdn.tailwindcss.com"></script>
<script
src="https://cdn.jsdelivr.net/gh/huggingface/hub-js-utils/share-canvas.js"
type="module"
></script>
<script type="module">
const MODEL_BASEURL =
"https://huggingface.co/lmz/candle-yolo-v8/resolve/main/";
const MODELS = {
yolov8n: {
model_size: "n",
url: "yolov8n.safetensors",
},
yolov8s: {
model_size: "s",
url: "yolov8s.safetensors",
},
yolov8m: {
model_size: "m",
url: "yolov8m.safetensors",
},
yolov8l: {
model_size: "l",
url: "yolov8l.safetensors",
},
yolov8x: {
model_size: "x",
url: "yolov8x.safetensors",
},
yolov8n_pose: {
model_size: "n",
url: "yolov8n-pose.safetensors",
},
yolov8s_pose: {
model_size: "s",
url: "yolov8s-pose.safetensors",
},
yolov8m_pose: {
model_size: "m",
url: "yolov8m-pose.safetensors",
},
yolov8l_pose: {
model_size: "l",
url: "yolov8l-pose.safetensors",
},
yolov8x_pose: {
model_size: "x",
url: "yolov8x-pose.safetensors",
},
};
const COCO_PERSON_SKELETON = [
[4, 0], // head
[3, 0],
[16, 14], // left lower leg
[14, 12], // left upper leg
[6, 12], // left torso
[6, 5], // top torso
[6, 8], // upper arm
[8, 10], // lower arm
[1, 2], // head
[1, 3], // right head
[2, 4], // left head
[3, 5], // right neck
[4, 6], // left neck
[5, 7], // right upper arm
[7, 9], // right lower arm
[5, 11], // right torso
[11, 12], // bottom torso
[11, 13], // right upper leg
[13, 15], // right lower leg
];
// init web worker
const yoloWorker = new Worker("./yoloWorker.js", { type: "module" });
let hasImage = false;
//add event listener to image examples
document.querySelector("#image-select").addEventListener("click", (e) => {
const target = e.target;
if (target.nodeName === "IMG") {
const href = target.src;
drawImageCanvas(href);
}
});
//add event listener to file input
document.querySelector("#file-upload").addEventListener("change", (e) => {
const target = e.target;
if (target.files.length > 0) {
const href = URL.createObjectURL(target.files[0]);
drawImageCanvas(href);
}
});
// add event listener to drop-area
const dropArea = document.querySelector("#drop-area");
dropArea.addEventListener("dragenter", (e) => {
e.preventDefault();
dropArea.classList.add("border-blue-700");
});
dropArea.addEventListener("dragleave", (e) => {
e.preventDefault();
dropArea.classList.remove("border-blue-700");
});
dropArea.addEventListener("dragover", (e) => {
e.preventDefault();
});
dropArea.addEventListener("drop", (e) => {
e.preventDefault();
dropArea.classList.remove("border-blue-700");
const url = e.dataTransfer.getData("text/uri-list");
const files = e.dataTransfer.files;
if (files.length > 0) {
const href = URL.createObjectURL(files[0]);
drawImageCanvas(href);
} else if (url) {
drawImageCanvas(url);
}
});
document.querySelector("#clear-btn").addEventListener("click", () => {
drawImageCanvas();
});
function drawImageCanvas(imgURL) {
const canvas = document.querySelector("#canvas");
const canvasResult = document.querySelector("#canvas-result");
canvasResult
.getContext("2d")
.clearRect(0, 0, canvas.width, canvas.height);
const ctx = canvas.getContext("2d");
ctx.clearRect(0, 0, canvas.width, canvas.height);
document.querySelector("#share-btn").classList.add("invisible");
document.querySelector("#clear-btn").classList.add("invisible");
document.querySelector("#detect").disabled = true;
hasImage = false;
canvas.parentElement.style.height = "auto";
if (imgURL && imgURL !== "") {
const img = new Image();
img.crossOrigin = "anonymous";
img.onload = () => {
canvas.width = img.width;
canvas.height = img.height;
ctx.drawImage(img, 0, 0);
canvas.parentElement.style.height = canvas.offsetHeight + "px";
hasImage = true;
document.querySelector("#detect").disabled = false;
document.querySelector("#clear-btn").classList.remove("invisible");
};
img.src = imgURL;
}
}
async function classifyImage(
imageURL, // URL of image to classify
modelID, // ID of model to use
modelURL, // URL to model file
modelSize, // size of model
confidence, // confidence threshold
iou_threshold, // IoU threshold
updateStatus // function receives status updates
) {
return new Promise((resolve, reject) => {
yoloWorker.postMessage({
imageURL,
modelID,
modelURL,
modelSize,
confidence,
iou_threshold,
});
yoloWorker.addEventListener("message", (event) => {
if ("status" in event.data) {
updateStatus(event.data.status);
}
if ("error" in event.data) {
reject(new Error(event.data.error));
}
if (event.data.status === "complete") {
resolve(event.data);
}
});
});
}
// add event listener to detect button
document.querySelector("#detect").addEventListener("click", async () => {
if (!hasImage) {
return;
}
const modelID = document.querySelector("#model").value;
const modelURL = MODEL_BASEURL + MODELS[modelID].url;
const modelSize = MODELS[modelID].model_size;
const confidence = parseFloat(
document.querySelector("#confidence").value
);
const iou_threshold = parseFloat(
document.querySelector("#iou_threshold").value
);
const canvasInput = document.querySelector("#canvas");
const canvas = document.querySelector("#canvas-result");
canvas.width = canvasInput.width;
canvas.height = canvasInput.height;
const scale = canvas.width / canvas.offsetWidth;
const ctx = canvas.getContext("2d");
ctx.drawImage(canvasInput, 0, 0);
const imageURL = canvas.toDataURL();
const results = await await classifyImage(
imageURL,
modelID,
modelURL,
modelSize,
confidence,
iou_threshold,
updateStatus
);
const { output } = results;
ctx.lineWidth = 1 + 2 * scale;
ctx.strokeStyle = "#3c8566";
ctx.fillStyle = "#0dff9a";
const fontSize = 14 * scale;
ctx.font = `${fontSize}px sans-serif`;
for (const detection of output) {
// check keypoint for pose model data
let xmin, xmax, ymin, ymax, label, confidence, keypoints;
if ("keypoints" in detection) {
xmin = detection.xmin;
xmax = detection.xmax;
ymin = detection.ymin;
ymax = detection.ymax;
confidence = detection.confidence;
keypoints = detection.keypoints;
} else {
const [_label, bbox] = detection;
label = _label;
xmin = bbox.xmin;
xmax = bbox.xmax;
ymin = bbox.ymin;
ymax = bbox.ymax;
confidence = bbox.confidence;
}
const [x, y, w, h] = [xmin, ymin, xmax - xmin, ymax - ymin];
const text = `${label ? label + " " : ""}${confidence.toFixed(2)}`;
const width = ctx.measureText(text).width;
ctx.fillStyle = "#3c8566";
ctx.fillRect(x - 2, y - fontSize, width + 4, fontSize);
ctx.fillStyle = "#e3fff3";
ctx.strokeRect(x, y, w, h);
ctx.fillText(text, x, y - 2);
if (keypoints) {
ctx.save();
ctx.fillStyle = "magenta";
ctx.strokeStyle = "yellow";
for (const keypoint of keypoints) {
const { x, y } = keypoint;
ctx.beginPath();
ctx.arc(x, y, 3, 0, 2 * Math.PI);
ctx.fill();
}
ctx.beginPath();
for (const [xid, yid] of COCO_PERSON_SKELETON) {
//draw line between skeleton keypoitns
if (keypoints[xid] && keypoints[yid]) {
ctx.moveTo(keypoints[xid].x, keypoints[xid].y);
ctx.lineTo(keypoints[yid].x, keypoints[yid].y);
}
}
ctx.stroke();
ctx.restore();
}
}
});
function updateStatus(statusMessage) {
const button = document.querySelector("#detect");
if (statusMessage === "detecting") {
button.disabled = true;
button.classList.add("bg-blue-700");
button.classList.remove("bg-blue-950");
button.textContent = "Predicting...";
} else if (statusMessage === "complete") {
button.disabled = false;
button.classList.add("bg-blue-950");
button.classList.remove("bg-blue-700");
button.textContent = "Predict";
document.querySelector("#share-btn").classList.remove("invisible");
}
}
document.querySelector("#share-btn").addEventListener("click", () => {
shareToCommunity(
"lmz/candle-yolo",
"Candle + YOLOv8",
"YOLOv8 with [Candle](https://github.com/huggingface/candle)",
"canvas-result",
"share-btn"
);
});
</script>
</head>
<body class="container max-w-4xl mx-auto p-4">
<main class="grid grid-cols-1 gap-8 relative">
<span class="absolute text-5xl -ml-[1em]"> 🕯️ </span>
<div>
<h1 class="text-5xl font-bold">Candle YOLOv8</h1>
<h2 class="text-2xl font-bold">Rust/WASM Demo</h2>
<p class="max-w-lg">
This demo showcases object detection and pose estimation models in
your browser using Rust/WASM. It utilizes
<a
href="https://huggingface.co/lmz/candle-yolo-v8"
target="_blank"
class="underline hover:text-blue-500 hover:no-underline"
>
safetensor's YOLOv8 models
</a>
and a WASM runtime built with
<a
href="https://github.com/huggingface/candle/"
target="_blank"
class="underline hover:text-blue-500 hover:no-underline"
>Candle </a
>.
</p>
<p>
To run pose estimation, select a yolo pose model from the dropdown
</p>
</div>
<div>
<label for="model" class="font-medium">Models Options: </label>
<select
id="model"
class="border-2 border-gray-500 rounded-md font-light"
>
<option value="yolov8n" selected>yolov8n (6.37 MB)</option>
<option value="yolov8s">yolov8s (22.4 MB)</option>
<option value="yolov8m">yolov8m (51.9 MB)</option>
<option value="yolov8l">yolov8l (87.5 MB)</option>
<option value="yolov8x">yolov8x (137 MB)</option>
<!-- Pose models -->
<option value="yolov8n_pose">yolov8n_pose (6.65 MB)</option>
<option value="yolov8s_pose">yolov8s_pose (23.3 MB)</option>
<option value="yolov8m_pose">yolov8m_pose (53 MB)</option>
<option value="yolov8l_pose">yolov8l_pose (89.1 MB)</option>
<option value="yolov8x_pose">yolov8x_pose (139 MB)</option>
</select>
</div>
<div>
<button
id="detect"
disabled
class="bg-gray-700 hover:bg-gray-800 text-white font-normal py-2 px-4 rounded disabled:bg-gray-300 disabled:cursor-not-allowed"
>
Predict
</button>
</div>
<!-- drag and drop area -->
<div class="relative">
<div class="py-1">
<button
id="clear-btn"
class="text-xs bg-white rounded-md disabled:opacity-50 flex gap-1 items-center ml-auto invisible"
>
<svg
class=""
xmlns="http://www.w3.org/2000/svg"
viewBox="0 0 13 12"
height="1em"
>
<path
d="M1.6.7 12 11.1M12 .7 1.6 11.1"
stroke="#2E3036"
stroke-width="2"
/>
</svg>
Clear image
</button>
</div>
<div
id="drop-area"
class="flex flex-col items-center justify-center border-2 border-gray-300 border-dashed rounded-xl relative aspect-video w-full overflow-hidden"
>
<div
class="flex flex-col items-center justify-center space-y-1 text-center"
>
<svg
width="25"
height="25"
viewBox="0 0 25 25"
fill="none"
xmlns="http://www.w3.org/2000/svg"
>
<path
d="M3.5 24.3a3 3 0 0 1-1.9-.8c-.5-.5-.8-1.2-.8-1.9V2.9c0-.7.3-1.3.8-1.9.6-.5 1.2-.7 2-.7h18.6c.7 0 1.3.2 1.9.7.5.6.7 1.2.7 2v18.6c0 .7-.2 1.4-.7 1.9a3 3 0 0 1-2 .8H3.6Zm0-2.7h18.7V2.9H3.5v18.7Zm2.7-2.7h13.3c.3 0 .5 0 .6-.3v-.7l-3.7-5a.6.6 0 0 0-.6-.2c-.2 0-.4 0-.5.3l-3.5 4.6-2.4-3.3a.6.6 0 0 0-.6-.3c-.2 0-.4.1-.5.3l-2.7 3.6c-.1.2-.2.4 0 .7.1.2.3.3.6.3Z"
fill="#000"
/>
</svg>
<div class="flex text-sm text-gray-600">
<label
for="file-upload"
class="relative cursor-pointer bg-white rounded-md font-medium text-blue-950 hover:text-blue-700"
>
<span>Drag and drop your image here</span>
<span class="block text-xs">or</span>
<span class="block text-xs">Click to upload</span>
</label>
</div>
<input
id="file-upload"
name="file-upload"
type="file"
class="sr-only"
/>
</div>
<canvas
id="canvas"
class="absolute pointer-events-none w-full"
></canvas>
<canvas
id="canvas-result"
class="absolute pointer-events-none w-full"
></canvas>
</div>
<div class="text-right py-2">
<button
id="share-btn"
class="bg-white rounded-md hover:outline outline-orange-200 disabled:opacity-50 invisible"
>
<img
src="https://huggingface.co/datasets/huggingface/badges/raw/main/share-to-community-sm.svg"
/>
</button>
</div>
</div>
<div>
<div class="flex gap-3 items-center" id="image-select">
<h3 class="font-medium">Examples:</h3>
<img
src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/candle/examples/sf.jpg"
class="cursor-pointer w-24 h-24 object-cover"
/>
<img
src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/candle/examples/bike.jpeg"
class="cursor-pointer w-24 h-24 object-cover"
/>
<img
src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/candle/examples/000000000077.jpg"
class="cursor-pointer w-24 h-24 object-cover"
/>
</div>
</div>
<div>
<div class="grid grid-cols-3 max-w-md items-center gap-3">
<label class="text-sm font-medium" for="confidence"
>Confidence Threshold</label
>
<input
type="range"
id="confidence"
name="confidence"
min="0"
max="1"
step="0.01"
value="0.25"
oninput="this.nextElementSibling.value = Number(this.value).toFixed(2)"
/>
<output
class="text-xs font-light px-1 py-1 border border-gray-700 rounded-md w-min"
>0.25</output
>
<label class="text-sm font-medium" for="iou_threshold"
>IoU Threshold</label
>
<input
type="range"
id="iou_threshold"
name="iou_threshold"
min="0"
max="1"
step="0.01"
value="0.45"
oninput="this.nextElementSibling.value = Number(this.value).toFixed(2)"
/>
<output
class="font-extralight text-xs px-1 py-1 border border-gray-700 rounded-md w-min"
>0.45</output
>
</div>
</div>
</main>
</body>
</html>