prediction works with images from napari, csv, and directories

src/allencell_ml_segmenter/core/event.py

@@ -32,6 +32,7 @@ class Event(Enum):
     ACTION_PREDICTION_POSTPROCESSING_AUTO_THRESHOLD = (
         "postprocessing_auto_threshold"
     )
+    ACTION_PREDICTION_INITIATED = "prediction_initiated"
     # Curation
     ACTION_CURATION_RAW_SELECTED = "curation_raw_selected"
     ACTION_CURATION_SEG1_SELECTED = "curation_seg1_selected"

src/allencell_ml_segmenter/prediction/file_input_widget.py

@@ -1,7 +1,7 @@
 from pathlib import Path
 from typing import List, Optional
 
-from napari.utils.events import Event
+from napari.utils.events import Event as NapariEvent
 from qtpy.QtCore import Qt
 from qtpy.QtWidgets import (
     QWidget,
@@ -15,6 +15,7 @@
     QSizePolicy,
 )
 
+from allencell_ml_segmenter.core.event import Event
 from allencell_ml_segmenter.widgets.input_button_widget import (
     InputButton,
     FileInputMode,
@@ -28,7 +29,7 @@
     CheckBoxListWidget,
 )
 
-from napari.viewer import Viewer
+from allencell_ml_segmenter.main.viewer import Viewer
 
 
 class PredictionFileInput(QWidget):
@@ -168,6 +169,11 @@ def __init__(self, model: PredictionModel, viewer: Viewer):
         grid_layout.setColumnStretch(1, 0)
 
         frame.layout().addLayout(grid_layout)
+        self._model.subscribe(
+            Event.ACTION_PREDICTION_INITIATED,
+            self,
+            self._set_selected_image_paths_from_napari
+        )
 
     def _on_screen_slot(self) -> None:
         """Prohibits usage of non-related input fields if top button is checked."""
@@ -184,11 +190,16 @@ def _from_directory_slot(self) -> None:
         self._browse_dir_edit.setEnabled(True)
         self._model.set_prediction_input_mode(PredictionInputMode.FROM_PATH)
 
-    def _update_layer_list(self, event: Optional[Event] = None) -> None:
+    def _update_layer_list(self, event: Optional[NapariEvent] = None) -> None:
         self._image_list.clear()
         for layer in self._viewer.get_layers():
             self._image_list.add_item(layer.name)
 
+    def _set_selected_image_paths_from_napari(self, event: Optional[Event] = None) -> None:
+        selected_indices: List[int] = self._image_list.get_checked_rows()
+        selected_paths: List[Path] = [self._viewer.get_layers()[i].source.path for i in selected_indices]
+        self._model.set_selected_paths(selected_paths)
+
     # TODO: replace with correct implementation and move to a service
     def map_input_file_directory_to_path_list(
         self, input_file_directory: str

src/allencell_ml_segmenter/prediction/model.py

@@ -26,13 +26,17 @@ def __init__(self):
         self._image_input_channel_index: int = None
         self._input_mode: PredictionInputMode = None
         self._output_directory: Path = None
+        self._selected_paths: List[Path] = None
 
         # state related to ModelInputWidget
         self._preprocessing_method: str = None
         self._postprocessing_method: str = None
         self._postprocessing_simple_threshold: float = None
         self._postprocessing_auto_threshold: str = None
 
+        # app state
+        self._prediction_running: bool = False
+
     def get_input_image_path(self) -> Path:
         """
         Gets list of paths to input images.
@@ -151,3 +155,20 @@ def set_prediction_input_mode(self, mode: PredictionInputMode) -> None:
 
     def get_prediction_input_mode(self) -> PredictionInputMode:
         return self._input_mode
+
+    def set_selected_paths(self, paths: List[Path]) -> None:
+        self._selected_paths = paths
+
+    def get_selected_paths(self) -> List[Path]:
+        return self._selected_paths
+
+    def set_prediction_running(self, is_prediction_running: bool) -> None:
+        # To run some setup for predictions
+        self.dispatch(Event.ACTION_PREDICTION_INITIATED)
+        # Shoots off a prediction run
+        self.dispatch(Event.PROCESS_PREDICTION)
+        self._prediction_running = is_prediction_running
+
+    def get_prediction_running(self) -> bool:
+        return self._prediction_running
+

src/allencell_ml_segmenter/prediction/view.py

@@ -91,7 +91,7 @@ def __init__(
 
     def run_btn_handler(self):
         # Just to test service for now.
-        self._prediction_model.dispatch(Event.PROCESS_PREDICTION)
+        self._prediction_model.set_prediction_running(True)
 
     def doWork(self):
         print("doWork - prediction")

src/allencell_ml_segmenter/services/prediction_service.py

@@ -13,7 +13,7 @@
 from pathlib import Path
 from typing import Union, Dict, List
 
-# from cyto_dl.api.model import CytoDLModel
+from cyto_dl.api.model import CytoDLModel
 from napari.utils.notifications import show_warning
 
 
@@ -74,14 +74,35 @@ def _predict_model(self, _: Event) -> None:
             )
             continue_prediction = False
 
-        # Create a CSV if user selects a folder of input images.
-        if (
-            self._prediction_model.get_prediction_input_mode()
+        # Check to see if user has selected an input mode
+        input_mode_selected: PredictionInputMode = self._prediction_model.get_prediction_input_mode()
+        if not input_mode_selected:
+            show_warning("Please select input images before running prediction.")
+            continue_prediction = False
+        elif (
+            input_mode_selected
             == PredictionInputMode.FROM_PATH
         ):
+            # User has selected a directory or a csv as input images
             input_path: Path = self._prediction_model.get_input_image_path()
             if input_path.is_dir():
+                # if input path selected is a directory, we need to manually write a CSV for cyto-dl
                 self.write_csv_for_inputs(list(input_path.glob("*.*")))
+            elif input_path.suffix != ".csv":
+                # This should not be possible with FileInputWidget- throw an error.
+                raise ValueError("Somehow the user has selected a non-csv/directory for input images. Should not be possible with FileInputWidget")
+        elif input_mode_selected == PredictionInputMode.FROM_NAPARI_LAYERS:
+            # User has selected napari image layers as input images
+            selected_paths_from_napari: List[Path] = self._prediction_model.get_selected_paths()
+            if len(selected_paths_from_napari) < 1:
+                # No image layers selected
+                show_warning("Please select at least 1 image from the napari layer before running prediction.")
+                continue_prediction = False
+            else:
+                # If user selects input images from napari, we need to manually write a csv for cyto-dl
+                self.write_csv_for_inputs(self._prediction_model.get_selected_paths())
+
+
 
         if continue_prediction:
             cyto_api: CytoDLModel = CytoDLModel()