test data added, channel input ilastik check

Author: rhornb

.gitignore

@@ -109,4 +109,8 @@ ENV/
 
 # IDE settings
 .vscode/
-.idea/
+.idea/
+
+*.zarr
+# Test data loaded via pooch
+/tests/data/10_5281_zenodo_14883998

create_env_script.sh

@@ -19,7 +19,7 @@ $COMMMAND run $COMMMAND create $LOCATION \
                                --override-channels \
                                -c pytorch \
                                -c ilastik-forge \
-                               -c conda-forge $PYTHON ilastik \
+                               -c conda-forge $PYTHON ilastik vigra \
                                --no-channel-priority --yes
 
 echo "Installing ilastik-tasks version $VERSION"

src/ilastik_tasks/ilastik_pixel_classification_segmentation.py

@@ -221,10 +221,17 @@ def ilastik_pixel_classification_segmentation(
     )
 
     # Setup Ilastik headless shell
-    # TODO: check if info about training mode can be retreived e.g. if model needs
-    # single or dual channel input and throw error if wrong input is provided
     shell = setup_ilastik(ilastik_model)
 
+    # Check model channel requirements
+    expected_channels = check_ilastik_model_channels(shell)
+    if expected_channels == 2 and channel2 is None:
+        raise ValueError(f"Ilastik model expects two channels as "
+                         "input but only one channel was provided")
+    elif expected_channels == 1 and channel2 is not None:
+        raise ValueError(f"Ilastik model expects 1 channel as "
+                         "input but two channels were provided")
+
     # Find channel index
     tmp_channel = get_channel_from_image_zarr(
         image_zarr_path=zarr_url,
@@ -526,6 +533,25 @@ def ilastik_pixel_classification_segmentation(
         )
 
 
+def check_ilastik_model_channels(shell) -> int:
+    """Check number of input channels expected by Ilastik model.
+    
+    Args:
+        shell: Initialized Ilastik shell with loaded model
+        
+    Returns:
+        int: Number of expected input channels
+    """
+    # Get dataSelection applet from workflow
+    data_selection = shell.workflow.dataSelectionApplet
+    
+    # Get slot info containing expected channels
+    slot_info = data_selection.topLevelOperator.DatasetRoles.value
+    
+    # Return number of expected channels
+    return len(slot_info)
+
+
 if __name__ == "__main__":
     from fractal_tasks_core.tasks._utils import run_fractal_task
 

tests/conftest.py

@@ -1,13 +1,10 @@
-# copied from 
-# https://github.com/fractal-analytics-platform/fractal-helper-tasks/blob/main/tests/conftest.py
-
 import os
 import shutil
 from pathlib import Path
 
 import pooch
 import pytest
-
+from devtools import debug
 
 @pytest.fixture(scope="session")
 def testdata_path() -> Path:
@@ -16,7 +13,7 @@ def testdata_path() -> Path:
 
 
 @pytest.fixture(scope="session")
-def zenodo_zarr(testdata_path: Path) -> list[str]:
+def zenodo_zarr_3d(testdata_path: Path) -> str:
     """
     This takes care of multiple steps:
 
@@ -25,16 +22,14 @@ def zenodo_zarr(testdata_path: Path) -> list[str]:
     3. Modify the Zarrs in tests/data, to add whatever is not in Zenodo
     """
 
-    # 1 Download Zarrs from Zenodo
-    DOI = "10.5281/zenodo.10257149"
+    # 1) Download Zarrs from Zenodo
+    DOI = "10.5281/zenodo.14883998"
     DOI_slug = DOI.replace("/", "_").replace(".", "_")
-    platenames = ["plate.zarr", "plate_mip.zarr"]
     rootfolder = testdata_path / DOI_slug
-    folders = [rootfolder / plate for plate in platenames]
+    folder = rootfolder / "AssayPlate_Greiner_CELLSTAR655090.zarr"
 
     registry = {
-        "20200812-CardiomyocyteDifferentiation14-Cycle1.zarr.zip": None,
-        "20200812-CardiomyocyteDifferentiation14-Cycle1_mip.zarr.zip": None,
+        "AssayPlate_Greiner_CELLSTAR655090.zarr.zip": None,
     }
     base_url = f"doi:{DOI}"
     POOCH = pooch.create(
@@ -45,31 +40,16 @@ def zenodo_zarr(testdata_path: Path) -> list[str]:
         allow_updates=False,
     )
 
-    for ind, file_name in enumerate(
-        [
-            "20200812-CardiomyocyteDifferentiation14-Cycle1.zarr",
-            "20200812-CardiomyocyteDifferentiation14-Cycle1_mip.zarr",
-        ]
-    ):
-        # 1) Download/unzip a single Zarr from Zenodo
-        file_paths = POOCH.fetch(
-            f"{file_name}.zip", processor=pooch.Unzip(extract_dir=file_name)
-        )
-        zarr_full_path = file_paths[0].split(file_name)[0] + file_name
-        folder = folders[ind]
-
-        # 2) Copy the downloaded Zarr into tests/data
-        if os.path.isdir(str(folder)):
-            shutil.rmtree(str(folder))
-        shutil.copytree(Path(zarr_full_path) / file_name, folder)
-    return [str(f) for f in folders]
+    file_name = "AssayPlate_Greiner_CELLSTAR655090.zarr"
+    # 2) Download/unzip a single Zarr from Zenodo
+    file_paths = POOCH.fetch(
+        f"{file_name}.zip", processor=pooch.Unzip(extract_dir=file_name)
+    )
+    zarr_full_path = file_paths[0].split(file_name)[0] + file_name
 
+    # 3) Copy the downloaded Zarr into tests/data
+    if os.path.isdir(str(folder)):
+        shutil.rmtree(str(folder))
+    shutil.copytree(Path(zarr_full_path) / file_name, folder)
+    return Path(folder)
 
-@pytest.fixture(scope="function")
-def tmp_zenodo_zarr(zenodo_zarr: list[str], tmpdir: Path) -> list[str]:
-    """Generates a copy of the zenodo zarrs in a tmpdir"""
-    zenodo_mip_path = str(tmpdir / Path(zenodo_zarr[1]).name)
-    zenodo_path = str(tmpdir / Path(zenodo_zarr[0]).name)
-    shutil.copytree(zenodo_zarr[0], zenodo_path)
-    shutil.copytree(zenodo_zarr[1], zenodo_mip_path)
-    return [zenodo_path, zenodo_mip_path]

tests/data/pixel_classifier_3D_dual_channel.ilp

@@ -2,28 +2,52 @@
 from pathlib import Path
 
 import pytest
+import zarr
 from devtools import debug
+
 from fractal_tasks_core.channels import ChannelInputModel
 
 from ilastik_tasks.ilastik_pixel_classification_segmentation import (
     ilastik_pixel_classification_segmentation,
 )
 
-def test_ilastik_pixel_classification_segmentation_task_2D_single_channel(tmp_zenodo_zarr: list[str]):
+# TODO: add 2D testdata
+
+@pytest.fixture(scope="function")
+def test_data_dir_3d(tmp_path: Path, zenodo_zarr_3d: list) -> str:
+    """
+    Copy a test-data folder into a temporary folder.
+    """
+    dest_dir = (tmp_path / "ilastik_data_3d").as_posix()
+    debug(zenodo_zarr_3d, dest_dir)
+    shutil.copytree(zenodo_zarr_3d, dest_dir)
+    return dest_dir
+
+
+def test_ilastik_pixel_classification_segmentation_task_3D(test_data_dir_3d):
     """
-    Test the 2D (e.g. MIP) ilastik_pixel_classification_segmentation task with single channel input.
+    Test the 3D ilastik_pixel_classification_segmentation task with dual channel input.
     """
-    ilastik_model = (Path(__file__).parent / "data/pixel_classifier_2D.ilp").as_posix()
-    zarr_url = f"{tmp_zenodo_zarr[1]}/B/03/0"
+    ilastik_model = (Path(__file__).parent / "data/pixel_classifier_3D_dual_channel.ilp").as_posix()
+    zarr_url = f"{test_data_dir_3d}/B/03/0"
 
     ilastik_pixel_classification_segmentation(
         zarr_url=zarr_url,
         level=0,
-        channel=ChannelInputModel(label="DAPI"),
-        channel2=None,
+        channel=ChannelInputModel(label="DAPI_2"),
+        channel2=ChannelInputModel(label="ECadherin_2"),
         ilastik_model=str(ilastik_model),
         output_label_name="test_label",
     )
-
-# TODO: add (2D) dual channel ilastik model for testing
-# TODO: add (3D) test data for dual channel testing - code works but could not yet find suitable test data
+    
+    # Test failing of task if model was trained with two channels 
+    # but only one is provided
+    with pytest.raises(ValueError):
+            ilastik_pixel_classification_segmentation(
+            zarr_url=zarr_url,
+            level=0,
+            channel=ChannelInputModel(label="DAPI_2"),
+            channel2=None,
+            ilastik_model=str(ilastik_model),
+            output_label_name="test_label_single_channel",
+        )