npy2bdv

A library for writing/reading HDF5/XML datasets of Fiji BigDataViewer/BigStitcher format as numpy arrays.

Installation

Run pip install npy2bdv in the command line. Use import npy2bdv in the code.

From source (development)

With mamba (or conda):

git clone https://github.com/nvladimus/npy2bdv.git
cd npy2bdv
mamba env create -f environment.yml
mamba activate npy2bdv

Or with pip only: pip install -e .[dev].

Run the tests from the repo root: python -m pytest test.

Releasing

Bump version in pyproject.toml, then publish a GitHub release tagged v<version>. The publish workflow runs the tests and uploads the package to PyPI. The docs are rebuilt and deployed to GitHub Pages on every push to master (docs workflow). To build them locally: pip install -e .[docs], then sphinx-build -b html docs/source docs/build/html.

Documentation

https://nvladimus.github.io/npy2bdv/: a user guide with code examples, and a detailed API reference.

Larger, timed examples are in the Jupyter notebook.

Supported H5/XML writing options

• pixel types uint16 (default), uint8, float32: BdvWriter(..., dtype='float32'). Types other than uint16 need an up-to-date Fiji (BigDataViewer-core >= 10.4.6).
• compression methods None, gzip, lzf (None by default).
• downsampling options:
  • any number of mipmap levels
  • computed via averaging, compatible with BigDataViewer/BigStitcher convention.
• user-defined block sizes for H5 storage (default 4,256,256)
• any number of time points, illuminations, channels, tiles, angles.
• arbitrary affine transformation for each individual view (e.g. translation, rotation, shear).
• arbitrary voxel calibration for each view, to account for spatial anisotropy.
• individual views can differ in dimensions, voxel size, voxel units, exposure time, and exposure units.
• missing views are labeled in XML automatically.
• support of additional meta-information:
  • camera properties: name, exposureTime, exposureUnits
  • microscope (name and version), user
• writing virtual stacks of arbitrary size, by plane or sub-stack. Handy when your stack is larger than your RAM.
  • virtual stacks can be written with multiple subsampling levels and compression.

New features

See Recent changes

Writing speed

Currently speeds between 150 - 320 MB/s were achieved with confidence. Speeds of up to 2300 MB/s reported in earlier versions are likely due to bugs in testing methods in notebook examples_h5writing.ipynb. More testing is underway.

Acknowledgements

This code was inspired by Talley Lambert's code and improved by generous input from Adam Glaser, VolkerH, Doug Shepherd, Peter H, Mike Taormina, Pradeep Rajasekhar and aaronkanzer.

To report issues or bugs please use the issues tool.

Citation

You can cite it like so

It will give me some dopamine and motivation to keep improving it :heart: