Merge pull request #16 from MPI-Dortmund/add-magic-gui

Add magicgui to requirements

pylintrc

@@ -0,0 +1 @@
+extension-pkg-whitelist=pyqt, pyqt5, pyqt5-qt5

setup.cfg

@@ -29,6 +29,7 @@ install_requires =
     matplotlib
     scipy
     napari-clusters-plotter >= 0.7.2
+    magicgui
 python_requires = >=3.10
 include_package_data = True
 package_dir =

src/napari_tomotwin/load_umap.py

@@ -6,13 +6,15 @@
 import numpy as np
 from matplotlib.patches import Circle
 from napari.utils import notifications
-
+from qtpy.QtCore import Qt
+from qtpy.QtGui import QGuiApplication # pylint: disable=E0611
+from typing import List
 plotter_widget: PlotterWidget = None
-circle: Circle = None
+circles: List[Circle] = []
 umap: pd.DataFrame
 
 def _draw_circle(data_coordinates, label_layer, umap):
-    global circle
+    global circles
     global plotter_widget
 
     label_layer.visible = 1
@@ -26,10 +28,15 @@ def _draw_circle(data_coordinates, label_layer, umap):
         center = umap_coordinates.values.tolist()[0]
     except IndexError:
         return
-
-    if circle is not None:
-        circle.remove()
+    modifiers = QGuiApplication.keyboardModifiers()
+    if modifiers == Qt.ShiftModifier:
+        pass
+    else:
+        for c in circles[::-1]:
+            c.remove()
+        circles = []
     circle = Circle(tuple(center), 0.5, fill=False, color='r')
+    circles.append(circle)
     plotter_widget.graphics_widget.axes.add_patch(circle)
     plotter_widget.graphics_widget.draw_idle()
 

src/napari_tomotwin/make_targets.py

@@ -15,8 +15,9 @@ def _get_medoid_embedding(embeddings: pd.DataFrame, max_embeddings: int = 50000)
     """
     if len(embeddings)>max_embeddings:
         # For samples more than 50k it's way to slow and memory hungry.
-        embeddings = embeddings.sample(max_embeddings)
         print(f"Your cluster size ({len(embeddings)}) is bigger then {max_embeddings}. Make a random sample to calculate medoid.")
+        embeddings = embeddings.sample(max_embeddings)
+
     only_emb = embeddings.drop(columns=["X", "Y", "Z", "filepath"], errors="ignore").astype(np.float32)
     distance_matrix=cdist(only_emb,only_emb,metric='cosine') # its not the cosine similarity, rather a distance (its 0 in case of same embeddings)
     medoid_index = np.argmin(np.sum(distance_matrix,axis=0))