Germinator AI

Supported Images

Upload a time-series of germination plate photos. The pipeline works best with:

• Standard digital camera photos (JPEG, PNG)
• Microscopy captures (TIFF, 8- or 16-bit)
• Web formats (WEBP)
• Top-down photos of seeds on filter paper or agar
• Consistent lighting and camera position between frames

Upload Methods

Four ways to load your data:

• Individual files — select or drag multiple images
• ZIP archive — drop a .zip containing your image series
• Time-lapse video — drop a single video file and choose how many evenly-spaced frames to extract
• Demo dataset — click below to load a sample Arabidopsis time-series

Files are sorted by name (use zero-padded numbering, e.g. frame_001.jpg or timestamps like 041224_1430.jpg).

How It Works

Germinator AI uses a multi-stage pipeline based on Joosen et al. (2010):

• Segment — excess-blue contrast separates seeds from background
• Calibrate — t₀ seed median defines the scale ruler
• Register — centroid voting aligns frames
• Track — geodesic BFS assigns each pixel to a seed identity
• Score — absorbing-state model detects germination onset

Output & Metrics

After analysis you get:

• Gmax — maximum germination percentage
• t₅₀ — time to 50% germination
• U₈₄₋₁₆ — germination uniformity
• AUC — area under curve
• Per-seed CSV export and full JSON report
• Interactive seed overlay map with germination state
• Hill-fit 4-parameter dose-response curve