Switched to voting system for classification

README.md

@@ -6,19 +6,21 @@
 
 # What is it?
 
-A Napari plugin that allows for the segmentation and detection of microglia on fluorescent images.
+A Napari plugin that allows for the segmentation and detection of microglia on 2D fluorescent images.
+Z-stacks are not handled.
 
 It consists in:
 - Segmenting the microglia with a UNet2D model.
 - Create classified bounding-boxes with a YOLOv5.
 - Using some morphology to extract metrics such as:
-    - The total length
-    - The length of the longest path
-    - The number of leaves
-    - The number of vertices
-    - Area of the convex hull
-    - Solidity/extent
-- We end-up with a TSV file containing all these metrics.
+    - The number of branches.
+    - The number of leaves (== end points).
+    - The number of vertices (== internal crossings).
+    - The mean branch length.
+    - The total length.
+    - The length of the longest path.
+
+- We end-up with a ".csv" (separated with tabulations instead of commas) file containing all these metrics.
 
 # 01. How to install/upgrade it?
 

changelog.md

@@ -0,0 +1,10 @@
+# Change Log
+
+## Version 1.1.4 (19 Jan. 2025)
+
+- Images normalization is now performed globally instead of locally to reduce the number of false positives.
+- The classification model was updated from the version 051 to the version 077.
+- The closing structuring element was changed from a diamong to a disk and its radius was reduced from 4 to 2.
+- For the reason just above, the default probability threshold was lowered from 40% to 15%.
+- The classification works now with a system of votes per label instead of taking every box above a certain score.
+- It is now possible to toggle the visibility of the "Garbage" class.

src/microglia_analyzer/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "1.0.4"
+__version__ = "1.1.4"
 
 import re