Command-line tool and python library for visualising small 3D datasets

image, volume, 3d, hdf5, n5, zarr
pip install smalldataviewer==0.10.0



Travis PyPI PyPI - Python Version

Simple matplotlib-based tool for viewing small amounts of 3D image data; helpful for debugging. Supports python 2.7 and 3.4+.

Adapted from this matplotlib recipe.

Small ssTEM Volume


pip install smalldataviewer[all]

The all installation includes all of these optional extras:

  • hdf5: HDF5 file support via h5py
  • img: volumetric and animated images via imageio
  • fits: FITS images via imageio[fits], which uses astropy
  • itk: ITK images via imageio[simpleitk], which uses SimpleITK

Support for N5 and zarr arrays is also available via z5py. This must be installed with conda (conda install -c conda-forge -c cpape z5py).


The DataViewer opens a matplotlib figure of the data volume.

  • Dimension 0 can be scrolled through with the mouse wheel
  • Dimension 1 is shown on the vertical axis
  • Dimension 2 is shown on the horizontal axis
  • Dimension 3, if it exists, is a colour tuple

As executable

Available as a command-line utility at smalldataviewer or sdv

usage: smalldataviewer [-h] [--version] [-i INTERNAL_PATH] [-t TYPE]
                       [-o ORDER] [-f OFFSET] [-s SHAPE] [-v] [-l]

positional arguments:
  path                  Path to file containing a 3D dataset

optional arguments:
  -h, --help            show this help message and exit
  --version             Print version information and exit
  -i INTERNAL_PATH, --internal_path INTERNAL_PATH
                        Internal path of dataset inside HDF5, N5, zarr or npz
                        file. If JSON, assumes the outer object is a dict, and
                        internal_path is the key of the array
  -t TYPE, --type TYPE  Dataset file type. Inferred from extension if not
  -o ORDER, --order ORDER
                        Order of non-channel axes for axis labelling purposes
                        (data is not transposed): dimension 0 will be scrolled
                        through, dimension 1 will be on the up-down axis,
                        dimension 2 will be on the left-right axis, and
                        dimension 3, if it exists, will be used as the colour
                        channels. Default "zyx".
  -f OFFSET, --offset OFFSET
                        3D offset of ROI from (0, 0, 0) in pixels, in the form
  -s SHAPE, --shape SHAPE
                        3D shape of ROI in pixels, in the form
  -v, --verbose         Increase logging verbosity
  -l, --label           Whether to treat images as a label volume


smalldataviewer my_data.hdf5 -i /my_group/my_volume

Note: because of the circumstances under which python holds file descriptors open, and under which matplotlib blocks, the executable form reads the data into memory in its entirety. If your data are too big for this, look at small chunks with the --offset (-f) and --shape (-s) options.

As library

import smalldataviewer as sdv

import numpy as np
data = np.random.random((30, 100, 100))
viewer = sdv.DataViewer(data)  # or

viewer2 = sdv.DataViewer.from_file(
    "my_data.npz", offset=(10, 20, 30), shape=(256, 512, 512), internal_path="volume"

reader = sdv.FileReader("my_cat_video.gif")
data2 =  # returns a numpy array
viewer3 = sdv.DataViewer(data2)

Note: FileReader (and by extension Dataviewer.from_file) reads the requested data from the file into memory. Passing an indexable representation of a file, like a numpy memmap or an hdf5 dataset, will not. However, you may need to copy it into memory for performance, or depending on the rest of your script.


Install a development environment (not including z5py) with make install-dev

Run tests in your current python environment with make test

Run tests against all supported python versions with make test-all

If you would like to add support for a new file type:

  1. Add to tests/common a function which creates such a file and returns whether it needs an internal path, and add it to file_constructors.

  2. Add to smalldataviewer.files.FileReader a method which reads such a file, returning a numpy array, and add a mapping from likely file extensions to a single file type in NORMALISED_TYPES (see existing methods for examples).

  3. Don't forget to specify any dependencies in smalldataviewer.ext, extras_require in, and requirements.txt