petrilya 0.0.1

AI colony counter for biology labs — drag, count, export.

Petrilya

AI colony counter for biology labs — drag, count, export.

Open-source desktop application that turns a phone photo of a Petri dish into a colony count, a CSV, and a one-page PDF report. No code, no cloud, no GPU required.

🌐 Website: petrilya.com · 🐙 Source: github.com/petrilya-app/petrilya-core

---

⚠️ Pre-alpha

Petrilya is being built in public. The current build runs the full UI (drag-and-drop, zoom, mask editing, batch processing, CSV/PDF/JSON export) against a mock segmentation engine — real Cellpose integration is blocked on cellpose.org model-weight hosting being available again. Until then, the mock generates plausible-looking colony detections so the rest of the pipeline can be tested.

If you need a working colony counter today, you probably want Cellpose directly. Petrilya's value-add is the no-code desktop wrapper, batch reports, and publication-ready outputs — coming soon.

---

Install (developer)

git clone https://github.com/petrilya-app/petrilya-core.git
cd petrilya-core
python -m venv .venv
.\.venv\Scripts\Activate.ps1     # Linux/macOS: source .venv/bin/activate
pip install -e ".[dev]"

Or, once published:

pip install petrilya

Run the GUI

petrilya-ui

CLI

petrilya path/to/dish.jpg --no-gpu
petrilya path/to/dish.jpg --gpu --diameter 30

---

What works today

• ✅ Drag-and-drop image loading (JPG, PNG, TIFF, BMP)
• ✅ Zoomable, pannable canvas (mouse wheel, Space-drag)
• ✅ Mask overlay with adjustable opacity (0–100%)
• ✅ Manual mask editing: erase a click, paint with brush
• ✅ Per-colony metrics: area, diameter, eccentricity, solidity
• ✅ Micrometers-per-pixel scale → results in μm² and μm
• ✅ Batch processing of an entire folder
• ✅ CSV export (per-colony rows)
• ✅ PDF report (preview + histogram + table)
• ✅ JSON manifest with SHA-256 and parameters for reproducibility
• ⏳ Real Cellpose model — pending cellpose.org server recovery

License

AGPL-3.0 for the open-source build. Commercial license available for organisations that need on-premise deployment without copyleft obligations — including GMP-relevant workflows. Contact [e-mail].

Acknowledgements

Built with Cellpose, PySide6, ONNX Runtime, scikit-image, and ReportLab.