Merge pull request #573 from AllenCell/feature/thresholding-live_images

Feature/thresholding live images

pyproject.toml

@@ -38,10 +38,10 @@ dependencies = [
     "npe2>=0.6.2",
     "numpy",
     "hydra-core==1.3.2",
-    "bioio",
+    "bioio==1.0.1",
     "tifffile>=2023.4.12",
     "watchdog",
-    "cyto-dl>=0.1.8",
+    "cyto-dl>=0.4.4",
     "scikit-image!=0.23.0",
 ]
 

src/allencell_ml_segmenter/_tests/core/test_file_input_model.py

@@ -1,8 +1,12 @@
 from pathlib import Path
 
+import allencell_ml_segmenter
 from allencell_ml_segmenter._tests.fakes.fake_subscriber import FakeSubscriber
 from allencell_ml_segmenter.core.event import Event
-from allencell_ml_segmenter.core.file_input_model import FileInputModel
+from allencell_ml_segmenter.core.file_input_model import (
+    FileInputModel,
+    InputMode,
+)
 
 
 def test_set_selected_paths_no_extract_channels() -> None:
@@ -98,3 +102,84 @@ def test_set_max_channels_dispatch() -> None:
 
     # Assert nothing happened
     dummy_subscriber.was_handled(Event.ACTION_FILEINPUT_MAX_CHANNELS_SET)
+
+
+def test_get_input_files_as_list_from_path() -> None:
+    """
+    Test to see if all paths from a directory are returned as a list
+    """
+    # ARRANGE
+    file_input_model: FileInputModel = FileInputModel()
+    file_input_model.set_input_mode(InputMode.FROM_PATH)
+    file_input_model.set_input_image_path(
+        Path(allencell_ml_segmenter.__file__).parent
+        / "_tests"
+        / "test_files"
+        / "img_folder"
+    )
+
+    # Act
+    files: list[Path] = file_input_model.get_input_files_as_list()
+
+    # Assert
+    assert len(files) == 5
+
+
+def test_get_input_files_as_list_from_viewer() -> None:
+    """
+    Test to see if all paths from viewer displayed images are returned as a list
+    """
+    # ARRANGE
+    file_input_model: FileInputModel = FileInputModel()
+    file_input_model.set_input_mode(InputMode.FROM_NAPARI_LAYERS)
+    fake_selected_paths: list[Path] = [Path("fake_path1"), Path("fake_path2")]
+    file_input_model.set_selected_paths(fake_selected_paths)
+
+    # Act
+    files: list[Path] = file_input_model.get_input_files_as_list()
+
+    # Assert
+    assert len(files) == 2
+    assert files == fake_selected_paths
+
+
+def test_get_input_files_as_list_from_no_directory_selected() -> None:
+    """
+    Test to see if an empty list is returned when no directory is selected
+    """
+    # ARRANGE
+    file_input_model: FileInputModel = FileInputModel()
+
+    # Act
+    files: list[Path] = file_input_model.get_input_files_as_list()
+
+    # Assert
+    assert len(files) == 0
+
+
+def test_get_input_files_as_list_from_no_selected_paths() -> None:
+    """
+    Test to see if an empty list is returned when no layers were selected
+    """
+    # ARRANGE
+    file_input_model: FileInputModel = FileInputModel()
+
+    # Act
+    files: list[Path] = file_input_model.get_input_files_as_list()
+
+    # Assert
+    assert len(files) == 0
+
+
+def test_get_input_files_as_list_from_no_selected_paths() -> None:
+    """
+    Test to see if an empty list is returned when no input mode is selected
+    """
+    # ARRANGE
+    file_input_model: FileInputModel = FileInputModel()
+
+    # Act
+    files: list[Path] = file_input_model.get_input_files_as_list()
+
+    # Assert
+    assert len(files) == 0

src/allencell_ml_segmenter/_tests/fakes/fake_viewer.py

@@ -24,9 +24,10 @@ def __init__(self, viewer: Optional[napari.Viewer] = None):
         self._shapes_layers: Dict[str, ShapesLayer] = {}
         self._labels_layers: Dict[str, LabelsLayer] = {}
         self._on_layers_change_fns: List[Callable] = []
+        self.threshold_inserted: Dict[str, np.ndarray] = {}
 
     def add_image(self, image: np.ndarray, name: str):
-        self._image_layers[name] = ImageLayer(name, None)
+        self._image_layers[name] = ImageLayer(name, path=None, data=image)
         self._on_layers_change()
 
     def get_image(self, name: str) -> Optional[ImageLayer]:
@@ -93,7 +94,7 @@ def contains_layer(self, name: str) -> bool:
 
     # not supporting in the fake because we will move away from this fn in the near future
     def get_layers(self) -> List[Layer]:
-        return []
+        return list(self._image_layers.values())
 
     def subscribe_layers_change_event(
         self, function: Callable[[NapariEvent], None]
@@ -103,3 +104,22 @@ def subscribe_layers_change_event(
     def _on_layers_change(self):
         for fn in self._on_layers_change_fns:
             fn(FakeNapariEvent())
+
+    def get_seg_layers(self) -> list[Layer]:
+        return [
+            layer
+            for layer in self._image_layers.values()
+            if layer.name.startswith("[seg]")
+        ]
+
+    def insert_threshold(
+        self, layer_name: str, img: np.ndarray, seg_layers: bool = False
+    ) -> None:
+        self.threshold_inserted[f"[threshold] {layer_name}"] = img
+
+    def get_layers_nonthreshold(self) -> list[Layer]:
+        return [
+            layer
+            for layer in self._image_layers.values()
+            if not layer.name.startswith("[threshold]")
+        ]

src/allencell_ml_segmenter/_tests/services/test_prediction_service.py

@@ -136,7 +136,7 @@ def test_build_overrides() -> None:
     assert overrides["train"] == False
     assert overrides["mode"] == "predict"
     assert overrides["task_name"] == "predict_task_from_app"
-    assert overrides["ckpt_path"] == str(
+    assert overrides["checkpoint.ckpt_path"] == str(
         Path(__file__).parent.parent
         / "main"
         / "experiments_home"

src/allencell_ml_segmenter/_tests/thresholding/test_thresholding_model.py

@@ -0,0 +1,54 @@
+import pytest
+
+from allencell_ml_segmenter._tests.fakes.fake_subscriber import FakeSubscriber
+from allencell_ml_segmenter.core.event import Event
+from allencell_ml_segmenter.thresholding.thresholding_model import (
+    ThresholdingModel,
+)
+
+
+@pytest.fixture
+def thresholding_model() -> ThresholdingModel:
+    model = ThresholdingModel()
+    return model
+
+
+def test_set_thresholding_value_dispatches_event(thresholding_model):
+    fake_subscriber: FakeSubscriber = FakeSubscriber()
+    thresholding_model.subscribe(
+        Event.ACTION_THRESHOLDING_VALUE_CHANGED,
+        fake_subscriber,
+        fake_subscriber.handle,
+    )
+
+    thresholding_model.set_thresholding_value(2)
+
+    assert fake_subscriber.was_handled(Event.ACTION_THRESHOLDING_VALUE_CHANGED)
+
+
+def test_set_autothresholding_enabled_dispatches_event(thresholding_model):
+    fake_subscriber: FakeSubscriber = FakeSubscriber()
+    thresholding_model.subscribe(
+        Event.ACTION_THRESHOLDING_AUTOTHRESHOLDING_SELECTED,
+        fake_subscriber,
+        fake_subscriber.handle,
+    )
+
+    thresholding_model.set_autothresholding_enabled(True)
+
+    assert fake_subscriber.was_handled(
+        Event.ACTION_THRESHOLDING_AUTOTHRESHOLDING_SELECTED
+    )
+
+
+def test_dispatch_save_thresholded_images(thresholding_model):
+    fake_subscriber: FakeSubscriber = FakeSubscriber()
+    thresholding_model.subscribe(
+        Event.ACTION_SAVE_THRESHOLDING_IMAGES,
+        fake_subscriber,
+        fake_subscriber.handle,
+    )
+
+    thresholding_model.dispatch_save_thresholded_images()
+
+    assert fake_subscriber.was_handled(Event.ACTION_SAVE_THRESHOLDING_IMAGES)

src/allencell_ml_segmenter/_tests/thresholding/test_thresholding_service.py

@@ -0,0 +1,101 @@
+import pytest
+import numpy as np
+
+from allencell_ml_segmenter.core.file_input_model import FileInputModel
+from allencell_ml_segmenter._tests.fakes.fake_experiments_model import (
+    FakeExperimentsModel,
+)
+from allencell_ml_segmenter.main.main_model import MainModel
+from allencell_ml_segmenter.thresholding.thresholding_model import (
+    ThresholdingModel,
+)
+from allencell_ml_segmenter.thresholding.thresholding_service import (
+    ThresholdingService,
+)
+from allencell_ml_segmenter.core.task_executor import SynchroTaskExecutor
+from allencell_ml_segmenter._tests.fakes.fake_viewer import FakeViewer
+
+
+@pytest.fixture
+def test_image():
+    """Create a small test image for thresholding."""
+    return np.array([[10, 20, 30], [40, 50, 60], [70, 80, 90]])
+
+
+def test_on_threshold_changed_non_prediction(test_image):
+    # ARRANGE
+    thresholding_model: ThresholdingModel = ThresholdingModel()
+    viewer: FakeViewer = FakeViewer()
+    thresholding_service: ThresholdingService = ThresholdingService(
+        thresholding_model,
+        FakeExperimentsModel(),
+        FileInputModel(),
+        MainModel(),
+        viewer,
+        task_executor=SynchroTaskExecutor.global_instance(),
+    )
+    viewer.add_image(test_image, name="test_layer")
+
+    # ACT set a threshold to trigger
+    thresholding_model.set_thresholding_value(50)
+
+    # Verify a segmentation layer is added
+    assert "[threshold] test_layer" in viewer.threshold_inserted
+    seg_data = viewer.threshold_inserted["[threshold] test_layer"]
+    assert np.array_equal(seg_data, (test_image > 50).astype(int))
+
+    # check if existing thresholds get updated
+    thresholding_model.set_thresholding_value(100)
+    assert len(viewer.get_layers()) == 1
+    seg_data = viewer.threshold_inserted["[threshold] test_layer"]
+    assert np.array_equal(seg_data, (test_image > 100).astype(int))
+
+
+def test_on_threshold_changed_non_prediction(test_image):
+    """
+    Test that the thresholding service does not add a threshold layer for a layer that is not a probability map
+    """
+    # ARRANGE
+    thresholding_model: ThresholdingModel = ThresholdingModel()
+    viewer: FakeViewer = FakeViewer()
+    main_model: MainModel = MainModel()
+    main_model.set_predictions_in_viewer(True)
+    thresholding_service: ThresholdingService = ThresholdingService(
+        thresholding_model,
+        FakeExperimentsModel(),
+        FileInputModel(),
+        main_model,
+        viewer,
+        task_executor=SynchroTaskExecutor.global_instance(),
+    )
+    # Only the [seg] layers below should produce a threshold layer
+    viewer.add_image(test_image, name="[raw] test_layer 1")
+    viewer.add_image(test_image, name="[seg] test_layer 1")
+    viewer.add_image(test_image, name="[raw] test_layer 2")
+    viewer.add_image(test_image, name="[seg] test_layer 2")
+    viewer.add_image(test_image, name="donotthreshold")
+
+    # ACT set a threshold to trigger
+    thresholding_model.set_thresholding_value(50)
+
+    # Verify a threshold layer is added for each seg layer
+    assert "[threshold] [seg] test_layer 1" in viewer.threshold_inserted
+    seg_data = viewer.threshold_inserted["[threshold] [seg] test_layer 1"]
+    assert np.array_equal(seg_data, (test_image > 50).astype(int))
+    assert "[threshold] [seg] test_layer 2" in viewer.threshold_inserted
+    seg_data = viewer.threshold_inserted["[threshold] [seg] test_layer 2"]
+    assert np.array_equal(seg_data, (test_image > 50).astype(int))
+    # verify that raw layers do not get thresholded
+    assert len(viewer.threshold_inserted) == 2
+
+    # verify existing threshold layers get updated correctly
+    thresholding_model.set_thresholding_value(100)
+    # Verify a threshold layer is added for each seg layer
+    assert "[threshold] [seg] test_layer 1" in viewer.threshold_inserted
+    seg_data = viewer.threshold_inserted["[threshold] [seg] test_layer 1"]
+    assert np.array_equal(seg_data, (test_image > 100).astype(int))
+    assert "[threshold] [seg] test_layer 2" in viewer.threshold_inserted
+    seg_data = viewer.threshold_inserted["[threshold] [seg] test_layer 2"]
+    assert np.array_equal(seg_data, (test_image > 100).astype(int))
+    # verify that raw layers do not get thresholded
+    assert len(viewer.threshold_inserted) == 2