Skip to content

VisCy

A deep learning pipeline for image-based phenotyping at single-cell resolution.1

  train and deploy virtual staining, segmentation, and self-supervised representation models on terabyte-scale imaging datasets.

What's inside

VisCy is a uv workspace monorepo: consisting of a core of independently versioned packages that compose into self-contained research applications.

  • viscy-data

    Data loading and Lightning DataModules for OME-Zarr microscopy datasets.

    Reference

  • viscy-models

    Neural network architectures - UNet, contrastive encoders, VAEs.

    Reference

  • viscy-transforms

    Image transforms tuned for virtual staining microscopy.

    Reference

  • viscy-utils

    Shared utilities across the other packages and applications.

    Reference

New here?

Start with Get started to install VisCy and then browse the packages overview for the live version of each package.

Interested in contrbuting?

See Contributing for environment setup, the pull request workflow, testing, and building the docs.

  1. VisCy is a blend of vision and cyto (cell) — computer vision for cell biology. ↩