viscy-transforms 0.1.0a0

Image transforms for virtual staining microscopy

viscy-transforms

Image transforms for virtual staining microscopy.

Part of the VisCy project.

Installation

From PyPI (when published)

pip install viscy-transforms

For development (from monorepo root)

# Using uv (recommended)
uv pip install -e packages/viscy-transforms

# Or via workspace sync
uv sync --package viscy-transforms

Usage

from viscy_transforms import NormalizeSampled, BatchedRandAffined

# Transforms follow MONAI dictionary transform pattern
# See documentation for full API reference

Features

• PyTorch-based image transforms optimized for microscopy data
• MONAI Dictionary transform compatibility for DataLoader pipelines
• Kornia-accelerated augmentations (affine, noise, blur)
• Specialized transforms for virtual staining workflows

Examples

To run the example notebook, install with the notebook extra:

pip install viscy-transforms[notebook]

See the batched transforms benchmark notebook for a comparison of batched GPU transforms vs standard MONAI transforms.

Dependencies

• torch>=2.4.1
• kornia
• monai>=1.4
• numpy

Documentation

In the works!

License

BSD-3-Clause - see LICENSE in repository root.