README.md

---
title: README
emoji: 🌍
colorFrom: gray
colorTo: yellow
sdk: static
pinned: false
---

<div style="background:#080d0f;border:1px solid #26363b;border-radius:20px;padding:28px;color:#f7efe9;">
  <div style="display:flex;align-items:center;gap:12px;margin-bottom:24px;">
    <img src="https://diffraction.agency/favicon.png" alt="" width="44" height="44" style="margin:0;" />
    <strong style="font-size:22px;letter-spacing:2px;color:#f7efe9;">DIFFRACTION</strong>
  </div>
  <p style="color:#65bfd1;font-size:11px;letter-spacing:2px;text-transform:uppercase;margin:0 0 12px;">Human observation · Physical AI · Applied community</p>
  <h2 style="font-family:Georgia,serif;font-size:40px;line-height:1.08;font-weight:400;color:#f7efe9;margin:0 0 18px;border:0;">Real tasks.<br /><span style="color:#65bfd1;">Inspectable training data.</span></h2>
  <p style="color:#c0cbd0;font-size:16px;line-height:1.7;">Diffraction builds human-demonstration datasets for robotics and physical AI. We connect task-specific capture, data processing and an applied AI community—with public samples you can inspect before starting a conversation.</p>
  <div style="display:flex;flex-wrap:wrap;gap:12px;margin-top:22px;">
    <a href="https://diffraction.agency/datasets/samples" style="display:inline-block;background:#65bfd1;color:#102026;padding:12px 18px;border-radius:10px;font-weight:700;text-decoration:none;">Explore the sample library →</a>
    <a href="https://diffraction.agency/contact" style="display:inline-block;background:#ffd21e;color:#241c00;padding:12px 18px;border-radius:10px;font-weight:700;text-decoration:none;">Discuss a custom dataset →</a>
  </div>
</div>

<h2>Start with the evidence</h2>
<p>Two task domains. Different signals. Each release includes its own provenance, quality notes and limitations.</p>

<div style="display:flex;flex-wrap:wrap;gap:18px;margin:20px 0;">
  <div style="flex:1 1 260px;min-width:0;border:1px solid #37474d;border-radius:16px;overflow:hidden;background:#10191e;color:#f7efe9;">
    <a href="https://huggingface.co/datasets/diffracting/egocentric-maintenance-sample" style="display:block;width:100%;"><img src="https://diffraction.agency/maintenance-sample-overview.webp" alt="Six chest-mounted appliance maintenance excerpts" style="display:block;width:100%;height:190px;object-fit:contain;margin:0;" /></a>
    <div style="padding:20px;">
      <p style="color:#65bfd1;font-size:11px;letter-spacing:1px;margin:0 0 10px;">MAINTENANCE / RGB</p>
      <h3 style="color:#f7efe9;margin:0 0 12px;font-size:23px;">Everyday tool use.</h3>
      <p style="color:#c0cbd0;line-height:1.6;">Tape removal, brushing and wiping. Six chest-mounted excerpts from one session, with source-frame mapping and approximate action intervals.</p>
      <p style="color:#c0cbd0;font-size:13px;">176.84 seconds · 5,308 frames · CC BY 4.0</p>
      <a href="https://huggingface.co/datasets/diffracting/egocentric-maintenance-sample" style="display:block;background:#ffd21e;color:#241c00;text-align:center;padding:12px;border-radius:9px;font-weight:700;text-decoration:none;margin:18px 0 12px;">🤗 Open maintenance dataset →</a>
      <a href="https://diffraction.agency/datasets/egocentric-maintenance-sample" style="color:#91d4e1;">Capture details &amp; evaluation notes →</a>
    </div>
  </div>
  <div style="flex:1 1 260px;min-width:0;border:1px solid #37474d;border-radius:16px;overflow:hidden;background:#10191e;color:#f7efe9;">
    <a href="https://huggingface.co/datasets/diffracting/egocentric-kitchen-sample" style="display:block;width:100%;"><img src="https://diffraction.agency/kitchen-sample-overview.webp" alt="Kitchen RGB, metric depth and confidence previews" style="display:block;width:100%;height:190px;object-fit:contain;background:#ffffff;margin:0;" /></a>
    <div style="padding:20px;">
      <p style="color:#65bfd1;font-size:11px;letter-spacing:1px;margin:0 0 10px;">KITCHEN / RGB-D</p>
      <h3 style="color:#f7efe9;margin:0 0 12px;font-size:23px;">A richer view of motion.</h3>
      <p style="color:#c0cbd0;line-height:1.6;">Onion preparation, pepper slicing and dicing. Three recordings with native RGB-D, calibration, device IMU and explicitly estimated annotations.</p>
      <p style="color:#c0cbd0;font-size:13px;">167.85 seconds · 10,070 frames · CC BY 4.0</p>
      <a href="https://huggingface.co/datasets/diffracting/egocentric-kitchen-sample" style="display:block;background:#ffd21e;color:#241c00;text-align:center;padding:12px;border-radius:9px;font-weight:700;text-decoration:none;margin:18px 0 12px;">🤗 Open kitchen dataset →</a>
      <a href="https://diffraction.agency/datasets/egocentric-kitchen-sample" style="color:#91d4e1;">Capture details &amp; evaluation notes →</a>
    </div>
  </div>
</div>

<p><strong>Know what you are using.</strong> These are small human-observation samples, not robot-control benchmarks. Estimated annotations are identified separately; neither release claims independently measured annotation accuracy or robot-policy gains. Maintenance has no depth or IMU. Consult each dataset card for available channels, missing data, grouping and license terms.</p>

<h2>From a sample to your specification</h2>
<p>Bring the task, environments, viewpoints and signals your project needs. We scope capture, processing, quality checks and delivery around an agreed brief. Controlled and distributed capture are configured to the project; robot-arm collection is in pilot buildout.</p>

<div style="padding:22px;border:1px solid #2b5059;border-radius:14px;background:#102329;color:#f7efe9;margin:20px 0;">
  <h3 style="color:#f7efe9;margin:0 0 10px;">Better data starts with a better brief.</h3>
  <p style="color:#c0cbd0;">Define tasks, measured and estimated signals, coverage and acceptance criteria with our practical buyer’s guide.</p>
  <a href="https://diffraction.agency/datasets/robotics-dataset-brief" style="display:inline-block;color:#102026;background:#65bfd1;border-radius:9px;padding:12px 16px;font-weight:700;text-decoration:none;">Read the guide &amp; get the free template →</a>
</div>

<h2>People are part of the pipeline</h2>
<p>The <a href="https://diffraction.agency/operators">Operators Lab</a> brings together builders, engineers, capture operators and physical-AI practitioners. Explore the community to learn, contribute and find project fit. Participation does not guarantee paid work.</p>
<p>Our wider work includes <a href="https://streamgenie.ai">creator content licensing through StreamGenie</a> and <a href="https://diffraction.agency/business-data">business data licensing</a>.</p>

<hr />
<p><strong>Follow this organization for new releases.</strong><br />
<a href="https://diffraction.agency">Website</a> · <a href="https://diffraction.agency/datasets/samples">Sample library</a> · <a href="https://diffraction.agency/operators">Operators Lab</a> · <a href="mailto:business@diffraction.agency">business@diffraction.agency</a></p>