5251 replace None type metadict content with 'none'

monai/data/image_reader.py

@@ -28,7 +28,16 @@
     orientation_ras_lps,
 )
 from monai.transforms.utility.array import EnsureChannelFirst
-from monai.utils import MetaKeys, SpaceKeys, deprecated, ensure_tuple, ensure_tuple_rep, optional_import, require_pkg
+from monai.utils import (
+    MetaKeys,
+    SpaceKeys,
+    TraceKeys,
+    deprecated,
+    ensure_tuple,
+    ensure_tuple_rep,
+    optional_import,
+    require_pkg,
+)
 
 if TYPE_CHECKING:
     import itk
@@ -131,7 +140,7 @@ def _copy_compatible_dict(from_dict: Dict, to_dict: Dict):
             datum = from_dict[key]
             if isinstance(datum, np.ndarray) and np_str_obj_array_pattern.search(datum.dtype.str) is not None:
                 continue
-            to_dict[key] = datum
+            to_dict[key] = str(TraceKeys.NONE) if datum is None else datum  # NoneType to string for default_collate
     else:
         affine_key, shape_key = MetaKeys.AFFINE, MetaKeys.SPATIAL_SHAPE
         if affine_key in from_dict and not np.allclose(from_dict[affine_key], to_dict[affine_key]):

monai/metrics/active_learning_metrics.py

@@ -15,6 +15,7 @@
 import torch
 
 from monai.metrics.utils import ignore_background
+from monai.utils import MetricReduction
 
 from .metric import Metric
 
@@ -135,7 +136,7 @@ def compute_variance(
 
     n_len = len(y_pred.shape)
 
-    if n_len < 4 and spatial_map is True:
+    if n_len < 4 and spatial_map:
         warnings.warn("Spatial map requires a 2D/3D image with N-repeats and C-channels")
         return None
 
@@ -149,16 +150,14 @@ def compute_variance(
     y_reshaped = torch.reshape(y_pred, new_shape)
     variance = torch.var(y_reshaped, dim=0, unbiased=False)
 
-    if spatial_map is True:
+    if spatial_map:
         return variance
 
-    elif spatial_map is False:
-        if scalar_reduction == "mean":
-            var_mean = torch.mean(variance)
-            return var_mean
-        elif scalar_reduction == "sum":
-            var_sum = torch.sum(variance)
-            return var_sum
+    if scalar_reduction == MetricReduction.MEAN:
+        return torch.mean(variance)
+    if scalar_reduction == MetricReduction.SUM:
+        return torch.sum(variance)
+    raise ValueError(f"scalar_reduction={scalar_reduction} not supported.")
 
 
 def label_quality_score(
@@ -196,13 +195,11 @@ def label_quality_score(
 
     abs_diff_map = torch.abs(y_pred - y)
 
-    if scalar_reduction == "none":
+    if scalar_reduction == MetricReduction.NONE:
         return abs_diff_map
 
-    elif scalar_reduction != "none":
-        if scalar_reduction == "mean":
-            lbl_score_mean = torch.mean(abs_diff_map, dim=list(range(1, n_len)))
-            return lbl_score_mean
-        elif scalar_reduction == "sum":
-            lbl_score_sum = torch.sum(abs_diff_map, dim=list(range(1, n_len)))
-            return lbl_score_sum
+    if scalar_reduction == MetricReduction.MEAN:
+        return torch.mean(abs_diff_map, dim=list(range(1, n_len)))
+    if scalar_reduction == MetricReduction.SUM:
+        return torch.sum(abs_diff_map, dim=list(range(1, n_len)))
+    raise ValueError(f"scalar_reduction={scalar_reduction} not supported.")

monai/visualize/occlusion_sensitivity.py

@@ -155,7 +155,8 @@ def gaussian_occlusion(x: torch.Tensor, mask_size, sigma=0.25) -> Tuple[torch.Te
             [GaussianSmooth(sigma=[b * sigma for b in spatial_shape]), Lambda(lambda x: -x), ScaleIntensity()]
         )
         # transform and add batch
-        mul: torch.Tensor = gaussian(kernel)[None]  # type: ignore
+        mul: torch.Tensor = gaussian(kernel)[None]
+
         return mul, 0
 
     @staticmethod

tests/test_pil_reader.py

@@ -64,6 +64,7 @@ def test_converter(self, data_shape, filenames, expected_shape, meta_shape):
                 Image.fromarray(test_image.astype("uint8")).save(filenames[i])
             reader = PILReader(converter=lambda image: image.convert("LA"))
             result = reader.get_data(reader.read(filenames, mode="r"))
+            self.assertEqual(result[1]["format"], "none")  # project-monai/monai issue#5251
             # load image by PIL and compare the result
             test_image = np.asarray(Image.open(filenames[0]).convert("LA"))