feat: migrate unet to torch, fix segservice

Author: henriquenunez

pyproject.toml

@@ -1,27 +1,13 @@
 [project]
 name = "partaker"
 version = "0.1.0"
-description = ""
+description = "Bacterial fluorescence analysis Tool for fluorescence microscopy"
 authors = [{ name = "Henrique Hiram" }]
 requires-python = ">=3.11,<3.12"
 readme = "README.md"
 license = { text = "LGPL" }
 dependencies = [
     "pyside6 (==6.7.2)",
-    # --- >>> PLATFORM-SPECIFIC TENSORFLOW (THIS BLOCK IS THE FIX) <<<
-    # Windows, Linux: regular tensorflow
-    "tensorflow (>=2.13.0,<2.16.0) ; platform_system != 'Darwin' and platform_system != 'Windows'",
-    # Mac ARM (Apple Silicon)
-    "tensorflow-macos (==2.15) ; platform_system == 'Darwin' and platform_machine == 'arm64'",
-    # Mac Intel (if you ever use Intel Mac)
-    "tensorflow-macos (==2.15) ; platform_system == 'Darwin' and platform_machine != 'arm64'",
-    # Windows: Intel/AMD optimized
-    "tensorflow-intel (>=2.13.0,<2.16.0) ; platform_system == 'Windows'",
-    # Tensorflow IO GCS filesystem for non-Mac-ARM
-    "tensorflow-io-gcs-filesystem (>=0.23.1) ; platform_machine != 'arm64' or platform_system != 'Darwin'",
-    # (Old) Windows: keep compatibility version
-    "tensorflow-io-gcs-filesystem (<0.32.0) ; platform_system == 'Windows'",
-    # --- >>> END OF PLATFORM-SPECIFIC BLOCK <<<
     "polars (>=1.30.0,<2.0.0)",
     "nd2 (>=0.10.3,<0.11.0)",
     "opencv-python (>=4.11.0.86,<5.0.0.0)",
@@ -63,3 +49,25 @@ dev = [
 
 [tool.hatch.build.targets.wheel]
 packages = ["src/nd2_analyzer"]
+
+[tool.ruff]
+target-version = "py312"
+
+[tool.ruff.lint]
+select = [
+  "E",    # pycodestyle errors
+  "F",    # pyflakes (undefined names, unused imports, etc.)
+  "W",    # pycodestyle warnings
+  "B",    # bugbear (common bugs and design issues)
+  "I",    # isort (import sorting)
+  "UP",   # pyupgrade (modernize syntax)
+  "RUF",  # ruff-specific rules
+  "N",    # pep8-naming
+  "ASYNC", # async pitfalls
+]
+ignore = ["E501"]  # line too long (ruff format handles this)
+
+[tool.ruff.format]
+quote-style = "double"
+indent-style = "space"
+

src/nd2_analyzer/analysis/metrics_service.py

@@ -46,19 +46,13 @@ def __init__(self):
             self._batch_size = 1000
             self._pending_count = 0
 
-            pub.subscribe(self.compute_metrics_at_frame, "image_ready")
+            pub.subscribe(self.compute_metrics_at_frame, "frame_segmented")
             self._initialized = True
 
     @timing_decorator("compute_metrics_at_frame")
     def compute_metrics_at_frame(
-        self, image: np.ndarray, time, position, channel, mode
+        self, labeled_frame: np.ndarray, time, position, channel
     ):
-        if mode != "segmented":
-            return
-
-        # labeled_frame = segmentation_cache[time, position, 0] # TODO: ensure segmentationservice always returns labeled
-        labeled_frame = image
-
         chan_n = ImageData.get_instance().channel_n
         mcherry_frame = yfp_frame = None
         if chan_n == 3:
@@ -112,18 +106,19 @@ def update_frame_metrics(self, batch_data: list):
     """
     Computes the metrics for each labeled cell from the segmentation
         # Uses regionprops to get geometrical features
-    
         # For fluorescence analysis:
         # Takes each segmented cell
         # 1. calculate physical metrics
         # 2. identify which fluorescence channel it is from
         # 3. write the actual fluorescence value
     """
 
-    def calculate_cell_metrics(_frame: TLFrame):
+    @classmethod
+    def calculate_cell_metrics(cls, _frame: TLFrame):
         cells = regionprops(_frame.labeled_phc)
         batch_data = []
 
+        print(f"shape phc {_frame.labeled_phc.shape} mcherry {_frame.mcherry.shape}")
         # Check for the fluorescence in any channel, if not, -1 and 0 fluorescence
         # mcherry can be None, same for yfp
         back_fluo_mcherry = (
@@ -141,6 +136,7 @@ def calculate_cell_metrics(_frame: TLFrame):
             fluorescence_level = 0.0
             has_fluorescence = False
 
+        logging.info(f"[calculate_cell_metrics] has {len(cells)} cells")
         for cell in cells:
             cell_id = cell.label
 
@@ -177,15 +173,21 @@ def calculate_cell_metrics(_frame: TLFrame):
             if has_fluorescence:
                 # If only mcherry has fluo
                 if back_fluo_mcherry != -1 and back_fluo_yfp == -1:
-                    mcherry_fluo = round(_frame.mcherry[_frame.labeled_phc == cell_id].mean(), 4)
+                    mcherry_fluo = round(
+                        _frame.mcherry[_frame.labeled_phc == cell_id].mean(), 4
+                    )
                     fluorescence_channel = 1
                     fluorescence_level = mcherry_fluo
 
                 # If both have fluorescence, compare them
                 elif back_fluo_mcherry != -1 and back_fluo_yfp != -1:
                     # Select the region corresponding to the cell in the frames
-                    mcherry_fluo = round(_frame.mcherry[_frame.labeled_phc == cell_id].mean(), 4)
-                    yfp_fluo = round(_frame.yfp[_frame.labeled_phc == cell_id].mean(), 4)
+                    mcherry_fluo = round(
+                        _frame.mcherry[_frame.labeled_phc == cell_id].mean(), 4
+                    )
+                    yfp_fluo = round(
+                        _frame.yfp[_frame.labeled_phc == cell_id].mean(), 4
+                    )
                     if back_fluo_mcherry != -1 and back_fluo_yfp != -1:
                         if (mcherry_fluo / back_fluo_mcherry) > (
                             yfp_fluo / back_fluo_yfp