Spaces:
Running
Running
File size: 4,556 Bytes
dcc5ee3 7bd1cc6 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 | <!doctype html>
<html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><meta name="color-scheme" content="dark"><title>ICP Lab · Dev Kumar</title><link rel="stylesheet" href="style.css"></head>
<body><main class="shell">
<header class="top"><a class="brand" href="https://github.com/devkumar-projects/Classical-Computer-Vision-Toolkit/tree/main/3d-point-cloud-registration-icp" target="_blank" rel="noopener"><span class="brand-mark">DK</span><span>DEV KUMAR <small>ROBOTICS & AI</small></span></a><a class="source-link" href="https://github.com/devkumar-projects/Classical-Computer-Vision-Toolkit/tree/main/3d-point-cloud-registration-icp" target="_blank" rel="noopener">Source project <span aria-hidden="true">↗</span></a></header>
<section class="intro"><div><p class="eyebrow"><span></span> INTERACTIVE COMPUTER VISION</p><h1>Watch point clouds<br><em>find their place.</em></h1><p class="lead">Explore iterative closest point (ICP), one iteration at a time. Adjust a synthetic 3D scan and see how a rigid alignment evolves.</p></div><div class="intro-side"><span>01 / ALIGNMENT</span><p>NEAREST NEIGHBOURS<br>RIGID TRANSFORMATION<br>CONVERGENCE</p></div></section>
<section class="workspace" aria-label="Interactive ICP demo">
<div class="visual panel"><div class="panel-head"><div><span class="section-kicker">LIVE VIEW</span><h2>Rigid registration</h2></div><div class="legend"><span><i class="dot target"></i>Reference</span><span><i class="dot moving"></i>Moving scan</span></div></div>
<div class="canvas-wrap"><canvas id="cloud" aria-label="Rotatable three-dimensional view of the reference and moving point clouds"></canvas><div class="canvas-note">DRAG TO ROTATE VIEW <span>•</span> SYNTHETIC DATA</div></div>
<div class="stats"><div><span>ITERATIONS</span><strong id="iterations">0</strong></div><div><span>NEAREST RMSE</span><strong id="rmse">—</strong></div><div><span>INITIAL RMSE</span><strong id="initial">—</strong></div><div><span>POINTS / CLOUD</span><strong id="points">230</strong></div></div>
</div>
<aside class="controls panel"><div class="control-head"><span class="section-kicker">EXPERIMENT</span><h2>Set the scene.</h2><p>Move the scan, then align it to the reference.</p></div>
<label for="rotation"><span>Rotation <output id="rotationVal">18°</output></span><input id="rotation" type="range" min="-35" max="35" step="1" value="18"></label>
<label for="shift"><span>Translation <output id="shiftVal">0.24</output></span><input id="shift" type="range" min="0" max="0.55" step="0.01" value="0.24"></label>
<label for="noise"><span>Noise σ <output id="noiseVal">0.00</output></span><input id="noise" type="range" min="0" max="0.12" step="0.005" value="0"></label>
<div class="buttons"><button id="play" class="primary">▶ Run ICP</button><button id="step">Step</button><button id="reset" aria-label="Reset experiment">Reset</button></div>
<div class="run-status" role="status" aria-live="polite"><i id="statusDot"></i><span id="status">Ready to align</span></div>
<div class="chart-head"><span>CONVERGENCE</span><span>NEAREST-NEIGHBOUR RMSE</span></div><canvas id="chart" aria-label="Chart of nearest-neighbour root mean square error by iteration"></canvas>
<p class="footnote">Units are arbitrary. ICP is a local method: symmetry, noise and initial pose can lead to incorrect matches or local minima.</p>
</aside>
</section>
<section class="explain"><div><span class="section-kicker">HOW IT WORKS</span><h2>Three steps.<br><em>Repeated.</em></h2></div><div class="explain-grid"><article><b>01</b><h3>Find neighbours</h3><p>For each moving point, select its closest point in the reference cloud.</p></article><article><b>02</b><h3>Fit the pose</h3><p>Estimate the rigid rotation and translation that reduce the paired distances.</p></article><article><b>03</b><h3>Iterate</h3><p>Transform the moving scan and recompute correspondences until improvement slows.</p></article></div></section>
<footer><span>DEV KUMAR · OPEN ROBOTICS & COMPUTER VISION</span><div><a href="https://www.kaggle.com/code/devkumarkaggle/3d-point-cloud-registration-with-icp" target="_blank" rel="noopener">Kaggle notebook ↗</a><a href="https://github.com/devkumar-projects/Classical-Computer-Vision-Toolkit/tree/main/3d-point-cloud-registration-icp" target="_blank" rel="noopener">GitHub project ↗</a></div></footer>
</main><script type="module" src="app.js"></script></body></html>
|