<div class="hero-badge">⚠ Research use only &mdash; not a medical device</div>
<h1>Segment cells<br><em>on your device</em></h1>
<p class="hero-sub">
  CellSeg runs the Cellpose cyto3 model entirely on your Android phone —
  no cloud, no account, no data upload required for local inference.
</p>
<div class="hero-actions">
  <a class="btn btn-primary btn-lg" href="/download">&#8659; Download APK</a>
  <a class="btn btn-outline btn-lg" href="https://github.com/lynchaos/cellseg" target="_blank" rel="noopener noreferrer">GitHub ↗</a>
</div>

<div class="section-title">
  <h2>Everything you need, in your pocket</h2>
  <p>From image capture to quantitative metrics — entirely offline.</p>
</div>
<div class="features-grid">

  <div class="feature-card">
    <div class="feature-icon">🧫</div>
    <h3>On-device inference</h3>
    <p>The 26 MB Cellpose cyto3 ONNX (FP16) model runs fully on-device via
       ONNX Runtime 1.19 with XNNPACK acceleration. No internet required.</p>
  </div>

  <div class="feature-card">
    <div class="feature-icon">☁</div>
    <h3>Cloud fallback</h3>
    <p>Optionally route to your own Hugging Face Gradio Space for
       higher-resolution inference or when device resources are limited.
       Your images and API token stay private.</p>
  </div>

  <div class="feature-card">
    <div class="feature-icon">📷</div>
    <h3>Camera &amp; TIFF import</h3>
    <p>Capture directly with the phone camera or import TIFF files
       from a microscopy export. Brightfield, phase-contrast, and
       fluorescence channels supported.</p>
  </div>

  <div class="feature-card">
    <div class="feature-icon">📊</div>
    <h3>Quantitative metrics</h3>
    <p>Cell count, confluence %, mean/median cell area, density
       (cells/cm²), and area histograms — computed per run and
       exportable as CSV.</p>
  </div>

  <div class="feature-card">
    <div class="feature-icon">📋</div>
    <h3>Batch processing</h3>
    <p>Queue multiple images for sequential segmentation. Export a
       combined metrics CSV for all runs with one tap.</p>
  </div>

  <div class="feature-card">
    <div class="feature-icon">🔒</div>
    <h3>Privacy-first</h3>
    <p>Zero analytics, no ads, no crash reporters. Your HuggingFace
       token is stored in EncryptedSharedPreferences. All files live
       in app-private storage.</p>
  </div>

</div>

<p class="built-on-label">Built on open source</p>
<div class="built-on-row">
  <span class="built-on-item">Cellpose cyto3</span>
  <span class="built-on-item">ONNX Runtime 1.19</span>
  <span class="built-on-item">Jetpack Compose</span>
  <span class="built-on-item">Hilt</span>
  <span class="built-on-item">Room</span>
  <span class="built-on-item">CameraX 1.4</span>
</div>