Vizarr is a minimal, purely client-side program for viewing Zarr-based images. It is built with Viv and exposes a Python API using the imjoy-rpc, allowing users to programatically view multiplex and multiscale images from within a Jupyter Notebook. The ImJoy plugin registers a codec for Python zarr.Array and zarr.Group objects, enabling Viv to securely request chunks lazily via Zarr.js. This means that other valid zarr-python stores can be viewed remotely with Viv, enabling flexible workflows when working with large datasets.

We created Vizarr to enhance interactive multimodal image alignment using the wsireg library. We describe a rapid workflow where comparison of registration methods as well as visual verification of alignnment can be assessed remotely, leveraging high-performance computational resources for rapid image processing and Viv for interactive web-based visualization in a laptop computer. The Jupyter Notebook containing the workflow described in the manuscript can be found in multimodal_registration_vizarr.ipynb. For more information, please read our preprint doi:10.31219/osf.io/wd2gu.

Vizarr supports viewing 2D slices of n-Dimensional Zarr arrays, allowing users to choose a single channel or blended composites of multiple channels during analysis. It has special support for the developing OME-Zarr format for multiscale and multimodal images. Currently Viv supports u1, u2, u4, and f4 arrays, but contributions are welcome to support more np.dtypes!

The easiest way to get started with vizarr is to clone this repoe and open one of the example Jupyter Notebooks.

vizarr was built to support the registration use case above where multiple, pyramidal OME-Zarr images are viewed within a Jupyter Notebook. Support for other Zarr arrays is supported but not as well tested. More information regarding thew viewing of generic Zarr arrays can be found in in the example notebooks.