Merge pull request #55 from fractal-napari-plugins-collection/51_prediction_layer_fixes

Fix prediction/annotation layer issues

Author: jluethi

src/napari_feature_classifier/annotator_widget.py

@@ -136,16 +136,7 @@ def __init__(
         for layer in self._viewer.layers:
             if type(layer) == napari.layers.Labels and layer.name == "Annotations":
                 self._viewer.layers.remove(layer)
-        self._annotations_layer = self._viewer.add_labels(
-            self._last_selected_label_layer.data,
-            scale=self._last_selected_label_layer.scale,
-            name="Annotations",
-            translate=self._last_selected_label_layer.translate,
-        )
-        self._annotations_layer.editable = False
-
-        # Set the label selection to a valid label layer => Running into proxy bug
-        self._viewer.layers.selection.active = self._last_selected_label_layer
+        self.add_annotations_layer()
 
         # Class selection
         self.ClassSelection = ClassSelection  # pylint: disable=C0103
@@ -212,11 +203,26 @@ def selection_changed(self, event):
             self._save_destination.enabled = False
             self._class_selector.enabled = False
 
+    def add_annotations_layer(self):
+        self._annotations_layer = self._viewer.add_labels(
+            self._last_selected_label_layer.data,
+            scale=self._last_selected_label_layer.scale,
+            name="Annotations",
+            translate=self._last_selected_label_layer.translate,
+        )
+        self._annotations_layer.editable = False
+        # Set the label selection to a valid label layer
+        self._viewer.layers.selection.active = self._last_selected_label_layer
+
     def toggle_label(self, labels_layer, event):
         """
         Callback for when a label is clicked. It then updates the color of that
         label in the annotation layer.
         """
+        # If the annotations layer is missing, add it back
+        if "Annotations" not in [x.name for x in self._viewer.layers]:
+            self.add_annotations_layer()
+
         # Need to translate & scale position that event.position returns by the
         # label_layer scale.
         # If scale is (1, 1, 1), nothing changes

src/napari_feature_classifier/classifier_widget.py

@@ -223,6 +223,7 @@ def __init__(
         self._last_selected_label_layer = get_selected_or_valid_label_layer(
             viewer=self._viewer
         )
+
         # Initialize the classifier
         if classifier:
             self._classifier = classifier
@@ -248,17 +249,10 @@ def __init__(
             self._viewer, get_class_selection(class_names=self.class_names)
         )
 
-        # Handle existing predictions layer
         for layer in self._viewer.layers:
             if type(layer) == napari.layers.Labels and layer.name == "Predictions":
                 self._viewer.layers.remove(layer)
-        self._prediction_layer = self._viewer.add_labels(
-            self._last_selected_label_layer.data,
-            scale=self._last_selected_label_layer.scale,
-            name="Predictions",
-            translate=self._last_selected_label_layer.translate,
-        )
-        self._prediction_layer.contour = 2
+        self.add_prediction_layer()
 
         # Set the label selection to a valid label layer => Running into proxy bug
         self._viewer.layers.selection.active = self._last_selected_label_layer
@@ -295,11 +289,6 @@ def __init__(
         self._export_button.clicked.connect(self.export_results)
         self._viewer.layers.selection.events.changed.connect(self.selection_changed)
         self._init_prediction_layer(self._last_selected_label_layer)
-        # Whenever the label layer is clicked, hide the prediction layer
-        # (e.g. new annotations are made)
-        # self._last_selected_label_layer.mouse_drag_callbacks.append(
-        #     self.hide_prediction_layer
-        # )
 
     def run(self):
         """
@@ -349,6 +338,15 @@ def add_features_to_classifier(self):
                     dict_of_features[layer.name] = layer.features
         self._classifier.add_dict_of_features(dict_of_features)
 
+    def add_prediction_layer(self):
+        self._prediction_layer = self._viewer.add_labels(
+            self._last_selected_label_layer.data,
+            scale=self._last_selected_label_layer.scale,
+            name="Predictions",
+            translate=self._last_selected_label_layer.translate,
+        )
+        self._prediction_layer.contour = 2
+
     def make_predictions(self):
         """
         Make predictions for all relevant label layers and add them to the
@@ -398,17 +396,72 @@ def selection_changed(self):
                 viewer=self._viewer
             ):
                 self._last_selected_label_layer = self._viewer.layers.selection.active
-                self._init_prediction_layer(self._viewer.layers.selection.active)
-                # self._last_selected_label_layer.mouse_drag_callbacks.append(
-                #     self.hide_prediction_layer
-                # )
+                self._init_prediction_layer(
+                    self._viewer.layers.selection.active, ensure_layer_presence=False
+                )
                 self._update_export_destination(self._last_selected_label_layer)
 
-    def _init_prediction_layer(self, label_layer: napari.layers.Labels):
+    def reorder_layers(self):
+        """Reorders layers if needed to ensure Annotation & Prediction layers
+        are above the currently selected label layer.
+        """
+        # Get the current order of layers
+        all_layers = list(self._viewer.layers)
+
+        # Determine the indices of the layers if they exist
+        indices_to_move = []
+
+        # Find the index of "Prediction" layer if it exists
+        if "Predictions" in self._viewer.layers:
+            indices_to_move.append(self._viewer.layers.index("Predictions"))
+
+        # Find the index of "Annotation" layer if it exists
+        if "Annotations" in self._viewer.layers:
+            indices_to_move.append(self._viewer.layers.index("Annotations"))
+
+        # Find the index of the reference_label_layer
+        if self._last_selected_label_layer.name in self._viewer.layers:
+            indices_to_move.append(
+                self._viewer.layers.index(self._last_selected_label_layer.name)
+            )
+
+        # Calculate the new order of layer indices
+        remaining_indices = [
+            i for i in range(len(all_layers)) if i not in indices_to_move
+        ]
+        remaining_indices.reverse()
+        new_order = indices_to_move + remaining_indices
+        new_order.reverse()
+
+        # Reorder the layers using the move_multiple function
+        self._viewer.layers.move_multiple(new_order)
+
+    def _init_prediction_layer(
+        self, label_layer: napari.layers.Labels, ensure_layer_presence: bool = True
+    ):
         """
         Initialize the prediction layer and reset its data (to fit the input
-        label_layer) and its colormap
+        label_layer) and its colormap.
+        ensure_layer_presence creates the Predictions layer if it doesn't exist
+        yet and triggers layer reordering.
         """
+        # Ensure that prediction layer exists
+        if (
+            "Predictions" not in [x.name for x in self._viewer.layers]
+            and ensure_layer_presence
+        ):
+            self.add_prediction_layer()
+        if ensure_layer_presence:
+            # Ensure correct layer order: This sometimes fails with weird
+            # EmitLoopError & IndexError that should be ignored
+            try:
+                self.reorder_layers()
+            except:  # noqa
+                pass
+
+        # Ensure that prediction layer is above the current label layer
+        self._last_selected_label_layer
+
         # Check if the predict column already exists in the layer.features
         if "prediction" not in label_layer.features:
             unique_labels = np.unique(label_layer.data)[1:]
@@ -448,12 +501,6 @@ def _init_prediction_layer(self, label_layer: napari.layers.Labels):
                 cmap=get_colormap(),
             )
 
-    # def hide_prediction_layer(self, labels_layer, event):
-    #     """
-    #     Hide the prediction layer
-    #     """
-    #     self._prediction_layer.visible = False
-
     def get_relevant_label_layers(self):
         relevant_label_layers = []
         required_columns = [self._label_column, self._roi_id_colum]
@@ -613,12 +660,19 @@ def load(self):
         with open(clf_path, "rb") as f:  # pylint: disable=C0103
             clf = pickle.load(f)
 
-        self._run_container = ClassifierRunContainer(
-            self._viewer,
-            clf,
-            classifier_save_path=clf_path,
-            auto_save=True,
-        )
+        try:
+            self._run_container = ClassifierRunContainer(
+                self._viewer,
+                clf,
+                classifier_save_path=clf_path,
+                auto_save=True,
+            )
+        except NotImplementedError:
+            napari_info(
+                "Create a label layer with a feature dataframe before loading "
+                "the classifier"
+            )
+            return
         self.clear()
         self.append(self._run_container)