wip

pyproject.toml

@@ -29,8 +29,10 @@ classifiers = [
 dependencies = [
   "cellpose",
   "confuse",
+  "pyimagej",
   "rich",
   "scikit-image",
+  "scyjava",
   "tqdm",
   "typer"
 ]

src/faim_wako_searchfirst/segment.py

@@ -72,3 +72,33 @@ def cellpose(
         **kwargs,
     )
     return mask
+
+
+def weka_classifier(
+    img,
+    classifier_path: Path,
+    logger=logging,
+):
+    """Apply a classifier.model file.
+
+    Use Trainable Weka Segmentation (via pyimagej)
+    to apply a trained classifier to the input image.
+    """
+    import imagej
+    from scyjava import jimport
+
+    logger.info("Initializing ImageJ...")
+    ij = imagej.init("sc.fiji:fiji:2.15.0")
+
+    WekaSegmentation = jimport("trainableSegmentation.WekaSegmentation")
+
+    logger.info("Starting Trainable Segmentation plugin...")
+    segmenter = WekaSegmentation(ij.py.to_imageplus(img))
+    logger.info(f"Loading classifier from {classifier_path}.")
+    segmenter.loadClassifier(str(classifier_path))
+    segmenter.applyClassifier(False)
+    # get result imp
+    # convert imp from_java
+    result = ij.py.from_java(segmenter.getClassifiedImage())
+    ij.dispose()
+    return label(result, background=1).astype(np.uint16)

tests/test_weka_classifier.py

@@ -0,0 +1,31 @@
+# SPDX-FileCopyrightText: 2024 Friedrich Miescher Institute for Biomedical Research (FMI), Basel (Switzerland)
+#
+# SPDX-License-Identifier: MIT
+"""Test segment.weka_classifier functionality."""
+from pathlib import Path
+
+import numpy as np
+import pytest
+from faim_wako_searchfirst.segment import weka_classifier
+from skimage.io import imread
+
+SAMPLE_IMAGE_NAME = "1649_1109_0003_Amp5-1_B_20070424_D11_w2_E2E74170-1089-4BE7-B8A3-6EC6852579B7.png"
+
+
+@pytest.fixture
+def _classifier_path():
+    return Path("tests/resources/classifier.model")
+
+
+@pytest.fixture
+def _sample_image():
+    return imread(Path("tests/resources") / SAMPLE_IMAGE_NAME)
+
+
+def test_weka_classifier(_sample_image, _classifier_path):
+    """Directly call weka_classifier."""
+    result = weka_classifier(_sample_image, _classifier_path)
+    assert result.shape == (442, 442)
+    assert np.count_nonzero(result == 1) == 3
+    assert np.count_nonzero(result == 2) == 8
+    assert np.count_nonzero(result == 3) == 831