New to PhenoMe or Python
Learn the basics with almost no code.
- Getting started (install + first script)
- 01 Interactive quickstart
- Learning paths
PhenoMe
Bring your images, pick any vision model, and turn embeddings into interpretable biology with no training or labels.
What is PhenoMe?
Section titled “What is PhenoMe?”PhenoMe is a reproducible phenotyping pipeline built around three ideas:
- Embeddings as the common representation. Any vision model that can describe images can be plugged in; all downstream analyses work on the resulting digital descriptions (vectors).
- Metadata as experimental context. Label images with conditions, doses, time points, plates, and wells.
- Classical properties to keep results easy to understand. Shape, intensity, texture, and quality metrics are computed alongside the embeddings and can be linked back to them to explain what the AI sees.
How it fits together
Section titled “How it fits together”
PhenoMe follows a modular 4-step workflow:
- Discover — Find your raw microscopy images and masks.
- Embeddings (A) — Preprocess images and run the vision model to extract visual fingerprints (digital descriptions of each image).
- Properties (B) — Extract explainable classical features (size, brightness, and texture).
- Analyze (C) — Merge both branches to compare images, visualize patterns (PCA/t-SNE/UMAP), explore interactively, and report.
The same four steps underpin every workflow. Everything else (interactive exploration, extension hooks, etc.) plugs into that backbone.
Choose your path
Section titled “Choose your path”Pick the card that matches you best. Each path ends at the same core idea: install, point at data, run notebooks or code, and understand your results.
Ready to Use Your Own Images
Clean and organize data, then plot results.
Developer or Integrator
Extend or embed PhenoMe into your pipeline.
Advanced Concepts
Deepen your understanding of embeddings and how to explain AI results.
Essential guides and references
Section titled “Essential guides and references”Understanding propertiesHow to read each built-in classical feature; preset bundles live under Select properties.
Function locationFind where a symbol lives in the codebase.
Report APIgenerate_report() parameters and HTML report output.