Update default batch size to 64 for inference

cellfinder/core/classify/cube_generator.py

@@ -40,7 +40,7 @@ def __init__(
         background_array: types.array,
         voxel_sizes: Tuple[int, int, int],
         network_voxel_sizes: Tuple[int, int, int],
-        batch_size: int = 32,
+        batch_size: int = 64,
         cube_width: int = 50,
         cube_height: int = 50,
         cube_depth: int = 20,
@@ -345,7 +345,7 @@ def __init__(
         signal_list: List[Union[str, Path]],
         background_list: List[Union[str, Path]],
         labels: Optional[List[int]] = None,  # only if training or validating
-        batch_size: int = 32,
+        batch_size: int = 64,
         shape: Tuple[int, int, int] = (50, 50, 20),
         channels: int = 2,
         classes: int = 2,

cellfinder/core/main.py

@@ -18,7 +18,7 @@ def main(
     trained_model: Optional[os.PathLike] = None,
     model_weights: Optional[os.PathLike] = None,
     model: model_type = "resnet50_tv",
-    batch_size: int = 32,
+    batch_size: int = 64,
     n_free_cpus: int = 2,
     network_voxel_sizes: Tuple[int, int, int] = (5, 1, 1),
     soma_diameter: int = 16,

cellfinder/napari/curation.py

@@ -54,7 +54,7 @@ def __init__(
         self.save_empty_cubes = save_empty_cubes
         self.max_ram = max_ram
         self.voxel_sizes = [5, 2, 2]
-        self.batch_size = 32
+        self.batch_size = 64
         self.viewer = viewer
 
         self.signal_layer = None

cellfinder/napari/detect/detect_containers.py

@@ -114,7 +114,7 @@ class ClassificationInputs(InputContainer):
     skip_classification: bool = False
     use_pre_trained_weights: bool = True
     trained_model: Optional[Path] = Path.home()
-    batch_size: int = 32
+    batch_size: int = 64
 
     def as_core_arguments(self) -> dict:
         args = super().as_core_arguments()

cellfinder/napari/detect/thread_worker.py

@@ -72,10 +72,10 @@ def detect_finished_callback(points: list) -> None:
         def classify_callback(batch: int) -> None:
             self.update_progress_bar.emit(
                 "Classifying cells",
-                # Default cellfinder-core batch size is 32. This seems to give
+                # Default cellfinder-core batch size is 64. This seems to give
                 # a slight underestimate of the number of batches though, so
                 # allow for batch number to go over this
-                max(self.npoints_detected // 32 + 1, batch + 1),
+                max(self.npoints_detected // 64 + 1, batch + 1),
                 batch + 1,
             )