Next-Shift / index.html
Lamapi's picture
Publish Next-Shift space
28b4298 verified
Raw History Blame Contribute Delete
2.52 kB
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link rel="icon" href="data:,">
<title>Next Shift</title>
<link rel="stylesheet" href="style.css">
</head>
<body>
<main>
<header>
<h1>Next Shift</h1>
<p>A digit recogniser trained on clean scans meets a new environment: grey, low-contrast, textured, rotated and blurred images. Label a few target images: the model fine-tunes its top layers on your labels in the browser, and its predictions on held-out target images update.</p>
<div class="badges"><span class="badge warn">Target environment is synthetic</span><span class="badge">Live in-browser adaptation</span><span class="badge">UCI digits (CC BY 4.0)</span><a class="badge" href="https://huggingface.co/thelamapi/Next-Shift">Model repo</a></div>
</header>
<div class="grid">
<section class="card">
<h2>Label target images</h2>
<div id="pool" class="chips" style="gap:4px"></div>
<div id="picker" style="margin-top:8px"></div>
<div class="row" style="margin-top:8px"><button id="auto" class="ghost">Add one dataset label per digit</button><button id="reset" class="ghost">Clear labels</button></div>
<label><input type="checkbox" id="hint"> Show dataset label as a hint</label>
</section>
<section class="card">
<h2>Held-out target accuracy</h2>
<div class="kpi">
<div><b id="k">0</b><span>your labels</span></div>
<div><b id="acc_ft">–</b><span>Next Shift</span></div>
<div><b id="acc_lin">–</b><span>frozen source features + linear head</span></div>
</div>
<p class="status" id="status">Loading…</p>
<canvas id="chart" style="margin-top:6px"></canvas><div id="cl"></div>
</section>
</div>
<section class="card" style="margin-top:12px">
<h2>Next Shift predictions on held-out target images</h2>
<div id="preds" style="display:grid;grid-template-columns:repeat(auto-fill,minmax(52px,1fr));gap:6px"></div>
</section>
<section class="card" style="margin-top:12px"><h2>Recorded benchmark</h2><img src="preview.png" alt="Accuracy vs labeled examples" style="width:100%;border-radius:8px"></section>
<footer>Source and target images come from the same UCI digits collection; the target "environment" is a synthetic visual transform.</footer>
</main>
<script src="nlrt.js"></script>
<script src="nlui.js"></script>
<script src="core.js"></script>
<script src="app.js"></script>
</body>
</html>