Segment cells
on your device
After a one-time ~25 MB model download, segmentation runs entirely on-device — no cloud, no account, no data upload required for inference.
Everything you need, in your pocket
From image capture to quantitative metrics — entirely offline.
On-device inference
The ~25 MB Cellpose cyto3 ONNX (FP32) model runs fully on-device via ONNX Runtime 1.19 with XNNPACK acceleration. No internet required.
Cloud fallback
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.
Camera & TIFF import
Capture directly with the phone camera or import TIFF files from a microscopy export. Brightfield, phase-contrast, and fluorescence channels supported.
Quantitative metrics
Cell count, confluence %, mean/median cell area, density (cells/cm²), and area histograms — computed per run and exportable as CSV.
Batch processing
Queue multiple images for sequential segmentation. Export a combined metrics CSV for all runs with one tap.
Privacy-first
Zero analytics, no ads, no crash reporters. Your HuggingFace token is stored in EncryptedSharedPreferences. All files live in app-private storage.
Built on open source