Merge pull request #29 from fractal-napari-plugins-collection/scikit-…

…learn-dependency

Update scikit learn dependency & test no roi_id code path

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = napari-feature-classifier
-version = 0.1.0
+version = 0.1.1
 author = Joel Luethi and Max Hess
 author_email = joel.luethi@uzh.ch
 url = https://github.com/fractal-napari-plugins-collection/napari-feature-classifier
@@ -43,7 +43,7 @@ install_requires =
     matplotlib
     magicgui
     pandas
-    scikit-learn
+    scikit-learn >= 1.2.2
     pandera
     xxhash
     hypothesis

src/napari_feature_classifier/_tests/test_classifier_widget.py

@@ -1,6 +1,8 @@
 """ Tests for classifier widget initialization"""
 import numpy as np
 import pandas as pd
+import pytest
+import os
 
 import imageio
 from pathlib import Path
@@ -31,17 +33,19 @@ def test_classifier_widgets_initialization_no_features_selected(make_napari_view
     assert classifier_widget._run_container is None
 
 
+
+features = make_features(np.unique(lbl_img_np)[1:], roi_id="ROI1", n_features=6)
+features_no_roi_id = features.drop(columns=["roi_id"])
 # make_napari_viewer is a pytest fixture that returns a napari viewer object
-# TODO: Verify the actual results of the classification
-def test_running_classification_through_widget(make_napari_viewer):
+@pytest.mark.parametrize("features", [features, features_no_roi_id])
+def test_running_classification_through_widget(features, make_napari_viewer):
     """
     Tests if the main widget launches
     """
     # make viewer and add an image layer using our fixture
     viewer = make_napari_viewer()
     label_layer = viewer.add_labels(lbl_img_np)
-    labels = np.unique(lbl_img_np)[1:]
-    label_layer.features = make_features(labels, roi_id="ROI1", n_features=6)
+    label_layer.features = features
 
     # Start init widget
     classifier_widget = ClassifierWidget(viewer)
@@ -66,7 +70,16 @@ def test_running_classification_through_widget(make_napari_viewer):
     # Run the classifier
     classifier_widget._run_container.run()
 
-    # TODO: Assert something
+    # Assert something that the layer is visible, predictions exist and are not NaN
     assert classifier_widget._run_container._prediction_layer.visible
     assert "prediction" in label_layer.features.columns
     assert pd.notna(label_layer.features["prediction"]).all().all()
+
+    # Check that the classifier file was saved
+    assert Path("lbl_img_np_classifier.clf").exists()
+
+    # Delete the classifier file (cleanup to avoid overwriting confirmation)
+    os.remove("lbl_img_np_classifier.clf")
+
+# TODO: Add a test to check the overwrite confirmations working correctly
+# For classifier files, for annotations and for exported predictions

src/napari_feature_classifier/classifier_widget.py

@@ -298,8 +298,12 @@ def run(self):
         self.add_features_to_classifier()
         try:
             self._classifier.train()
-        except ValueError:
-            napari_info("Not enough annotations made. Please make more annotations.")
+        except ValueError as e:
+            napari_info(
+                "Training failed. A typical reason are not having "
+                "enough annotations. \nThe error message was: "
+                f"{e}"
+            )
         else:
             self.make_predictions()
             self._prediction_layer.visible = True
@@ -502,7 +506,9 @@ def _update_export_destination(self, label_layer: napari.layers.Labels):
 
         """
         base_path = Path(self._export_destination.value).parent
-        self._export_destination.value = base_path / f"{label_layer.name}_predictions.csv"
+        self._export_destination.value = (
+            base_path / f"{label_layer.name}_predictions.csv"
+        )
 
     def export_results(self):
         """