Skip to content

Usage with Napari

Imaging Server Kit can generate Napari dock widgets for algorithms, algorithm collections, and algorithm servers.

Note

Install the Napari extra first: pip install "imaging-server-kit[napari]".

Opening a widget

From the Napari plugin menu

The Imaging Server Kit plugin adds a few entries to the Plugins > Imaging Server Kit menu:
Connect to server A widget that connects to an algorithm server
Demo algorithms The demo algorithms (same as serverkit demo napari)
Tool algorithms The built-in algorithms (same as serverkit tools napari)
Connect to QuPath The QuPath bridge

To start Napari with the Connect to server widget already open, run:

napari -w imaging-server-kit

From Python

You can pass an algorithm, an algorithm collection, or a client to sk.to_napari. This will add the widget to a new viewer and return that viewer:

import imaging_server_kit as sk

viewer = sk.to_napari(my_algo)

To add the widget to an existing viewer instead, pass it with viewer=:

import napari

viewer = napari.Viewer()
sk.to_napari(my_algo, viewer=viewer)

Tip

Calling napari.run() is needed in Python scripts to keep the viewer open (but not in Jupyter notebooks or IPython).

Widget overview

Napari widget

From top to bottom, the widget contains:

  • Server URL: the address of the algorithm server (used when connecting to a server).
  • Algorithm: a dropdown to select the algorithm (and access its 🌐 Doc page).
  • Samples: a dropdown to select and load samples into the viewer (only visible when there are samples).
  • Parameters: one input field per parameter, generated from the parameter annotations.
  • Tiled inference (when the algo has set tileable=True): run the algorithm tile-by-tile.
  • Run and ❌ Cancel, and a progress bar.

Parameter fields

Parameter layer Field
sk.Integer, sk.Float Spin box, using min, max, step, and default
sk.Bool Checkbox
sk.Choice Dropdown of the items
sk.String Text field
sk.Image, sk.Mask, sk.Points, sk.Boxes, sk.Vectors, sk.Paths, sk.Tracks Dropdown of the matching layers in the viewer
sk.Any, sk.Null Not shown

The name of a parameter is used as its label, and its description is shown as a tooltip.

When a parameter has auto_call=True, changing its value re-runs the algorithm.

Results in the viewer

Each output layer is shown in the viewer according to its type:

Output layer Shown as
sk.Image Image layer
sk.Mask Labels layer
sk.Points Points layer
sk.Boxes, sk.Paths Shapes layer
sk.Vectors Vectors layer
sk.Tracks Tracks layer
sk.Float, sk.Integer, sk.Bool, sk.String, sk.Choice Text overlay
sk.Notification Napari notification (info, warning, or error)
sk.Progress The widget's progress bar

Notice that when the algorithm runs again, layers with the same name are updated rather (they are not added again).

Extra keyword arguments of output layers are applied as properties of the Napari layer. For example, sk.Image(data, colormap="viridis") sets the colormap of the Napari Image layer.

Sending results to a viewer from Python

You can also run algorithms from Python and display the results in Napari:

  • algo.run(..., stack=viewer) adds the results to an existing viewer.
  • sk.convert(stack, to="napari") opens a new viewer showing a stack.

Usage in Napari plugins

sk.to_qwidget(algo, viewer) returns the widget without adding it to a viewer. This can be used to integrate Imaging Server Kit algorithms in other Napari plugins.