Stream live updates

Algorithms can send intermediate outputs before they finish running. This lets user interfaces such as Napari update in real time.

For example, you might stream:

• Frames from a camera, with a segmentation overlay or detection boxes.
• Progress notifications from a long-running task.
• Intermediate results of an iterative algorithm.

Yielding outputs

When a function decorated with @sk.algorithm contains yield statements, the yielded values are treated as intermediate outputs. The yielded values should be data layers, just like return values.

Layers with the same name are updated (rather than added), so only the last value of each layer remains in the final stack.

Here is an algorithm that progressively increases a threshold:

import time

import imaging_server_kit as sk
import numpy as np
import skimage.data

@sk.algorithm
def threshold_algo(image, steps=20):
    for k, threshold in enumerate(np.linspace(0, 255, steps)):  # Progressively increase the threshold
        mask = image > threshold
        yield sk.Mask(mask), sk.String(f"Threshold: {threshold}")  # <- Stream the outputs
        yield sk.Progress(k, max_val=steps)  # <- Update the progress bar
        time.sleep(0.5)

viewer = sk.to_napari(threshold_algo)
viewer.add_image(skimage.data.coins())

In Napari, the segmentation mask is updated as the threshold increases, and the progress bar of the widget fills up progressively.

Combining yield and return

An algorithm can yield intermediate outputs and still produce a final result with return:

import time

import imaging_server_kit as sk
import skimage.data

@sk.algorithm
def mini_pipeline(image, threshold=100):
    yield sk.Notification("Starting the processing")

    time.sleep(2)

    mask = image > threshold

    yield sk.Notification("Segmented the image"), sk.Mask(mask)

    time.sleep(2)

    fract = mask.sum() / mask.size

    return sk.Notification(f"Fraction of True pixels: {fract}")

viewer = sk.to_napari(mini_pipeline)
viewer.add_image(skimage.data.coins())

Here, the algorithm notifies the user that processing has started, then yields a segmentation mask, and finally reports a measurement.

Notifications and progress

• sk.Notification is used to show a message to the user. In Napari, it appears as a notification. You can set its level with level="info" (default), "warning", or "error".
• sk.Progress updates a progress bar, up to max_val.

See Data layers for the other layer types.