Add apoc pixel classifier segmentation

.github/workflows/test.yml

@@ -36,6 +36,22 @@ jobs:
       with:
         python-version: ${{ matrix.python-version }}
 
+    # # these libraries, along with pytest-xvfb (added in the `deps` in tox.ini),
+    # # enable testing on Qt on linux
+    # - name: Install Linux libraries
+    #   if: runner.os == 'Linux'
+    #   run: |
+    #     sudo apt-get install -y libdbus-1-3 libxkbcommon-x11-0 libxcb-icccm4 \
+    #       libxcb-image0 libxcb-keysyms1 libxcb-randr0 libxcb-render-util0 \
+    #       libxcb-xinerama0 libxcb-xinput0 libxcb-xfixes0
+
+    # # strategy borrowed from vispy for installing opengl libs on windows
+    # - name: Install Windows OpenGL
+    #   if: runner.os == 'Windows'
+    #   run: |
+    #     git clone --depth 1 https://github.com/pyvista/gl-ci-helpers.git
+    #     powershell gl-ci-helpers/appveyor/install_opengl.ps1
+
     - name: Install Hatch
       run: pip install --upgrade hatch
 

config_apoc.yml

@@ -0,0 +1,19 @@
+# SPDX-FileCopyrightText: 2024 Friedrich Miescher Institute for Biomedical Research (FMI), Basel (Switzerland)
+#
+# SPDX-License-Identifier: MIT
+
+# Required
+file_selection:  # criteria for file selection in case of multiple channels/slices per position
+    channel: C01
+process:  # choose method how to segment, filter, and sample the objects
+    segment: apoc_classifier
+    filter: [area]
+    sample: centers
+
+# Each subsequent section provides arguments to one of the methods defined in 'process'
+area:
+    min_area: 100
+    max_area: 1000000
+apoc_classifier:
+    classifier_path: "tests/resources/PixelClassifier.cl"
+    background_label: 2

pyproject.toml

@@ -27,8 +27,10 @@ classifiers = [
   "Programming Language :: Python :: Implementation :: PyPy",
 ]
 dependencies = [
+  "apoc",
   "cellpose",
   "confuse",
+  "pyopencl",
   "rich",
   "scikit-image",
   "tqdm",
@@ -45,11 +47,22 @@ Source = "https://github.com/fmi-faim/faim-wako-searchfirst"
 source = "vcs"
 
 [tool.hatch.envs.default]
+platforms = ["windows", "macos"]
 dependencies = [
   "pandas",
   "pytest",
   "pytest-cov",
 ]
+
+[tool.hatch.envs.linux]
+platforms = ["linux"]
+dependencies = [
+  "pandas",
+  "pocl-binary-distribution",
+  "pytest",
+  "pytest-cov",
+]
+
 [tool.hatch.envs.default.scripts]
 cov = "pytest --cov-report=term-missing --cov-config=pyproject.toml --cov=faim_wako_searchfirst --cov=tests {args}"
 no-cov = "cov --no-cov {args}"

src/faim_wako_searchfirst/segment.py

@@ -13,7 +13,6 @@
 from typing import Union
 
 import numpy as np
-from cellpose import models
 from scipy.ndimage import binary_fill_holes
 from skimage.filters import gaussian
 from skimage.measure import label
@@ -61,6 +60,8 @@ def cellpose(
 
     :return: a label image representing the detected objects
     """
+    from cellpose import models
+
     logger.info(f"Load cellpose model: {pretrained_model}")
     model: models.CellposeModel = models.CellposeModel(
         pretrained_model=pretrained_model,
@@ -72,3 +73,23 @@ def cellpose(
         **kwargs,
     )
     return mask
+
+
+def apoc_classifier(
+    img,
+    classifier_path: Path,
+    background_label: int = None,
+    logger=logging,
+):
+    """Apply a PixelClassifier.cl file.
+
+    Use apoc (https://github.com/haesleinhuepf/apoc)
+    to apply a trained classifier to the input image.
+    """
+    import apoc
+
+    clf = apoc.PixelClassifier(opencl_filename=classifier_path)
+    mask = clf.predict(image=img)
+    labeled_image, num_objects = label(mask, return_num=True, background=background_label)
+    logger.info(f"Found {num_objects} connected components.")
+    return labeled_image.astype(np.uint16)