Combine algorithms¶

Combining algorithms into a collection gives users access to several algorithms from a single interface, such as the Algorithm dropdown in Napari or QuPath, or a single server.

Creating a collection with `sk.combine`¶

Consider two separately defined algorithms:

import imaging_server_kit as sk

@sk.algorithm(
    name="threshold",
    description="Segment a grayscale image based on an intensity threshold.",
)
def threshold_algo(image, threshold=128):
    mask = image > threshold
    return sk.Mask(mask, name="Binary mask")

@sk.algorithm(
    name="foreground",
    description="Compute the fraction of positive pixels in a binary mask.",
)
def foreground_fract(mask):
    fract = mask.sum() / mask.size
    return sk.Float(fract, name="Foreground fraction")

Combine them into a collection with sk.combine:

multi_algo = sk.combine([threshold_algo, foreground_fract], name="segmentation-pipeline")

sk.to_napari(multi_algo)

Both algorithms are now available from the Algorithm dropdown in Napari.

Using a collection from Python¶

Algorithm collections have the same methods as standalone algorithms. Use the algorithm argument to select an algorithm by name:

import skimage.data

image = skimage.data.coins()

results = multi_algo.run(algorithm="threshold", image=image, threshold=100)

The list of algorithm names is available as multi_algo.algorithms. If algorithm is omitted, the first algorithm of the collection is used.

Serving a collection¶

Collections can be served like a single algorithm. On the server side:

import imaging_server_kit as sk
import skimage.data
from skimage.filters import threshold_otsu, threshold_li

@sk.algorithm(
    name="intensity-threshold",
    parameters={"threshold": sk.Integer(name="Threshold", min=0, max=255, default=128)},
    samples=[{"image": skimage.data.coins()}],
)
def threshold_algo(image, threshold):
    mask = image > threshold
    return sk.Mask(mask, name="Binary mask")

@sk.algorithm(
    name="automatic-threshold",
    parameters={"method": sk.Choice(name="Method", items=["otsu", "li"], default="otsu")},
    samples=[{"image": skimage.data.coins()}],
)
def auto_threshold(image, method):
    if method == "otsu":
        mask = image > threshold_otsu(image)
    elif method == "li":
        mask = image > threshold_li(image)
    return sk.Mask(mask, name="Binary mask")

threshold_algos = sk.combine([threshold_algo, auto_threshold], name="threshold-algos")

if __name__ == "__main__":
    sk.serve(threshold_algos)

Then, on the client side:

import imaging_server_kit as sk
import skimage.data

image = skimage.data.coins()

client = sk.Client("http://localhost:8000")

print(client.algorithms)  # ['intensity-threshold', 'automatic-threshold']

thresh_results = client.run(algorithm="intensity-threshold", image=image, threshold=30)
auto_results = client.run(algorithm="automatic-threshold", image=image, method="otsu")

Built-in algorithms¶

The Imaging Server Kit comes with a collection of common algorithms, sk.tools. It includes filters (Gaussian, median, Sobel, etc.), mask utilities (remove small objects, label, fill holes, etc.), math operations, thresholds (Otsu, manual), and more.

Open them in Napari from Plugins > Imaging Server Kit > Tool algorithms, from the command line with serverkit tools napari (or serve them with serverkit tools serve), or from Python:

import imaging_server_kit as sk

sk.to_napari(sk.tools)

Like any collection, sk.tools can be combined with your own algorithms, for example sk.combine([threshold_algo, *sk.tools.algorithms_dict.values()]).

Combine algorithms¶

Creating a collection with sk.combine¶

Using a collection from Python¶

Serving a collection¶

Built-in algorithms¶

Creating a collection with `sk.combine`¶