mlarray 0.0.32__tar.gz → 0.0.33__tar.gz

{mlarray-0.0.32 → mlarray-0.0.33}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mlarray
-Version: 0.0.32
+Version: 0.0.33
 Summary: Array format specialized for Machine Learning with Blosc2 backend and standardized metadata.
 Author-email: Karol Gotkowski <karol.gotkowski@dkfz.de>
 License: MIT

{mlarray-0.0.32 → mlarray-0.0.33}/docs/schema.md

@@ -97,15 +97,17 @@ This section stores precomputed global statistics for the array, which can be us
 ### bbox
 
 * **Description:** Bounding boxes for objects/regions in the image.
-* **Dataclass:** `MetaBbox` (single-key wrapper).
+* **Dataclass:** `MetaBbox`.
 * **Structure:** List of bboxes, each bbox is a list with length equal to image `ndims`,
   and each entry is `[min, max]`.
 
 Bounding boxes are stored in a normalized, axis-aligned representation that works across dimensionalities (2D, 3D, …). This is especially useful for detection-style workflows, ROI cropping, dataset summaries, and interactive visualization.
 
-| field  | type                            | description                                                          |
-| ------ | ------------------------------- | -------------------------------------------------------------------- |
-| bboxes | Optional[List[List[List[int]]]] | Bounding boxes shaped `[num_bboxes][ndims][2]` (min/max), ints only. |
+| field  | type                                            | description                                                                 |
+| ------ | ----------------------------------------------- | --------------------------------------------------------------------------- |
+| bboxes | Optional[List[List[List[Union[int, float]]]]]   | Bounding boxes shaped `[num_bboxes][ndims][2]` (min/max), ints or floats.   |
+| scores | Optional[List[Union[int, float]]]               | Optional confidence scores aligned with `bboxes`.                           |
+| labels | Optional[List[Union[str, int, float]]]          | Optional labels aligned with `bboxes`.                                      |
 
 ---
 

{mlarray-0.0.32 → mlarray-0.0.33}/examples/example_channel.py

@@ -1,7 +1,7 @@
 import numpy as np
 import os
 from pathlib import Path
-from mlarray import MLArray, Meta
+from mlarray import MLArray, Meta, MetaBbox
 import json
 
 
@@ -20,7 +20,7 @@ if __name__ == '__main__':
         os.remove(filepath)
 
     print("Initializing image...")
-    image = MLArray(array, spacing=spacing, origin=origin, direction=direction, channel_axis=channel_axis, meta=Meta(original=image_meta, bbox=bboxes))
+    image = MLArray(array, spacing=spacing, origin=origin, direction=direction, channel_axis=channel_axis, meta=Meta(original=image_meta, bbox=MetaBbox(bboxes)))
     print("Saving image...")
     image.save(filepath)
 

{mlarray-0.0.32 → mlarray-0.0.33}/examples/example_metadata_only.py

@@ -1,7 +1,7 @@
 import numpy as np
 import os
 from pathlib import Path
-from mlarray import MLArray, Meta
+from mlarray import MLArray, Meta, MetaBbox
 import json
 
 
@@ -19,7 +19,7 @@ if __name__ == '__main__':
         os.remove(filepath)
 
     print("Initializing image...")
-    image = MLArray(spacing=spacing, origin=origin, direction=direction, meta=Meta(original=image_meta, bbox=bboxes))
+    image = MLArray(spacing=spacing, origin=origin, direction=direction, meta=Meta(original=image_meta, bbox=MetaBbox(bboxes)))
     print("Saving image...")
     image.save(filepath)
 

{mlarray-0.0.32 → mlarray-0.0.33}/examples/example_open.py

@@ -1,7 +1,7 @@
 import numpy as np
 import os
 from pathlib import Path
-from mlarray import MLArray, Meta, MetaSpatial
+from mlarray import MLArray, Meta, MetaSpatial, MetaBbox
 import json
 
 
@@ -22,7 +22,7 @@ if __name__ == '__main__':
     image = MLArray.open(filepath, shape=array.shape, dtype=array.dtype, mmap='w+')
     print("Saving image...")
     image[...] = array
-    image.meta.copy_from(Meta(original=image_meta, spatial=MetaSpatial(spacing=spacing, origin=origin, direction=direction), bbox=bboxes))
+    image.meta.copy_from(Meta(original=image_meta, spatial=MetaSpatial(spacing=spacing, origin=origin, direction=direction), bbox=MetaBbox(bboxes)))
     image.meta.is_seg = True
     image.close()
 

{mlarray-0.0.32 → mlarray-0.0.33}/examples/example_save_load.py

@@ -1,8 +1,7 @@
 import numpy as np
 import os
 from pathlib import Path
-from mlarray import MLArray, Meta
-from mlarray.meta import MetaOriginal, MetaIsSeg
+from mlarray import MLArray, Meta, MetaBbox
 import json
 
 
@@ -20,7 +19,7 @@ if __name__ == '__main__':
         os.remove(filepath)
 
     print("Initializing image...")
-    image = MLArray(array, spacing=spacing, origin=origin, direction=direction, meta=Meta(original=image_meta, bbox=bboxes, is_seg=True))
+    image = MLArray(array, spacing=spacing, origin=origin, direction=direction, meta=Meta(original=image_meta, bbox=MetaBbox(bboxes), is_seg=True))
     print("Saving image...")
     image.save(filepath)
 

{mlarray-0.0.32 → mlarray-0.0.33}/mlarray/meta.py

@@ -575,34 +575,57 @@ class MetaStatistics(BaseMeta):
 
 
 @dataclass(slots=True)
-class MetaBbox(SingleKeyBaseMeta):
-    """Bounding boxes represented as per-dimension min/max pairs.
+class MetaBbox(BaseMeta):
+    """Bounding box metadata with optional scores and labels.
 
     Attributes:
         bboxes: List of bounding boxes with shape [n_boxes, ndims, 2], where
-            each inner pair is [min, max] for a dimension. Values must be ints.
+            each inner pair is [min, max] for a dimension. Values must be ints
+            or floats.
+        scores: Optional confidence scores aligned with bboxes (ints or floats).
+        labels: Optional labels aligned with bboxes. Each label may be a string,
+            int, or float.
     """
-    bboxes: Optional[List[List[List[int]]]] = None
+    bboxes: Optional[List[List[List[Union[int, float]]]]] = None
+    scores: Optional[List[Union[int, float]]] = None
+    labels: Optional[List[Union[str, int, float]]] = None
 
     def _validate_and_cast(self, **_: Any) -> None:
-        """Validate bounding box structure and types."""
-        if self.bboxes is None:
-            return
-
-        self.bboxes = _cast_to_list(self.bboxes, "meta.bbox.bboxes")
-
-        if not isinstance(self.bboxes, list):
-            raise TypeError("meta.bbox.bboxes must be a list")
-
-        for bbox in self.bboxes:
-            if not isinstance(bbox, list):
-                raise TypeError("meta.bbox.bboxes must be a list of lists")
-            for row in bbox:
-                if not isinstance(row, list) or len(row) != 2:
-                    raise ValueError("meta.bbox.bboxes rows must have length 2")
-                for v in row:
-                    if isinstance(v, bool) or not isinstance(v, int):
-                        raise TypeError("meta.bbox.bboxes must contain ints only")
+        """Validate bounding box structure and related fields."""
+        if self.bboxes is not None:
+            self.bboxes = _cast_to_list(self.bboxes, "meta.bbox.bboxes")
+
+            if not isinstance(self.bboxes, list):
+                raise TypeError("meta.bbox.bboxes must be a list")
+
+            for b_i, bbox in enumerate(self.bboxes):
+                if not isinstance(bbox, list):
+                    raise TypeError("meta.bbox.bboxes must be a list of lists")
+                for r_i, row in enumerate(bbox):
+                    if not isinstance(row, list) or len(row) != 2:
+                        raise ValueError("meta.bbox.bboxes rows must have length 2")
+                    for v in row:
+                        if isinstance(v, bool) or not isinstance(v, (float, int)):
+                            raise TypeError("meta.bbox.bboxes must contain ints or floats only")
+
+        if self.scores is not None:
+            self.scores = _cast_to_list(self.scores, "meta.bbox.scores")
+            _validate_float_int_list(self.scores, "meta.bbox.scores")
+
+        if self.labels is not None:
+            self.labels = _cast_to_list(self.labels, "meta.bbox.labels")
+            if not isinstance(self.labels, list):
+                raise TypeError("meta.bbox.labels must be a list")
+            for v in self.labels:
+                if isinstance(v, bool) or not isinstance(v, (str, int, float)):
+                    raise TypeError("meta.bbox.labels must contain only str, int, or float")
+
+        if self.bboxes is not None:
+            n = len(self.bboxes)
+            if self.scores is not None and len(self.scores) != n:
+                raise ValueError("meta.bbox.scores must have same length as bboxes")
+            if self.labels is not None and len(self.labels) != n:
+                raise ValueError("meta.bbox.labels must have same length as bboxes")
 
 
 @dataclass(slots=True)

{mlarray-0.0.32 → mlarray-0.0.33}/mlarray.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mlarray
-Version: 0.0.32
+Version: 0.0.33
 Summary: Array format specialized for Machine Learning with Blosc2 backend and standardized metadata.
 Author-email: Karol Gotkowski <karol.gotkowski@dkfz.de>
 License: MIT

{mlarray-0.0.32 → mlarray-0.0.33}/mlarray.egg-info/SOURCES.txt

@@ -34,6 +34,7 @@ mlarray.egg-info/dependency_links.txt
 mlarray.egg-info/entry_points.txt
 mlarray.egg-info/requires.txt
 mlarray.egg-info/top_level.txt
+tests/test_bboxes.py
 tests/test_metadata.py
 tests/test_optimization.py
 tests/test_usage.py

mlarray-0.0.33/tests/test_bboxes.py

@@ -0,0 +1,67 @@
+import tempfile
+import unittest
+from pathlib import Path
+
+import numpy as np
+
+from mlarray import MLArray
+from mlarray.meta import Meta, MetaBbox
+
+
+def _make_array(shape=(8, 16, 16), seed=0):
+    rng = np.random.default_rng(seed)
+    return rng.random(shape, dtype=np.float32)
+
+
+class TestMetaBbox(unittest.TestCase):
+    def test_bbox_accepts_ints_and_floats(self):
+        bbox = MetaBbox(bboxes=[[[0, 1.5], [2, 3], [4.0, 5]]])
+        self.assertEqual(bbox.bboxes, [[[0, 1.5], [2, 3], [4.0, 5]]])
+
+    def test_bbox_scores_and_labels(self):
+        bbox = MetaBbox(
+            bboxes=[[[0, 1], [2, 3]]],
+            scores=[0.9],
+            labels=["lesion"],
+        )
+        self.assertEqual(bbox.scores, [0.9])
+        self.assertEqual(bbox.labels, ["lesion"])
+
+    def test_bbox_scores_labels_length_mismatch(self):
+        with self.assertRaises(ValueError):
+            MetaBbox(bboxes=[[[0, 1], [2, 3]]], scores=[0.9, 0.8])
+        with self.assertRaises(ValueError):
+            MetaBbox(bboxes=[[[0, 1], [2, 3]]], labels=["a", "b"])
+
+    def test_bbox_labels_type_validation(self):
+        with self.assertRaises(TypeError):
+            MetaBbox(bboxes=[[[0, 1], [2, 3]]], labels=[True])
+
+    def test_bbox_casts_numpy_and_tuple_inputs(self):
+        bboxes = np.array([[[0, 1], [2, 3]]], dtype=np.int64)
+        bbox = MetaBbox(bboxes=bboxes, scores=(0.5,), labels=("a",))
+        self.assertEqual(bbox.bboxes, [[[0, 1], [2, 3]]])
+        self.assertEqual(bbox.scores, [0.5])
+        self.assertEqual(bbox.labels, ["a"])
+
+    def test_bbox_roundtrip_mlarray(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            array = _make_array()
+            meta = Meta(
+                bbox=MetaBbox(
+                    bboxes=[[[0, 1], [2, 3], [4, 5]]],
+                    scores=[0.7],
+                    labels=[1],
+                )
+            )
+            image = MLArray(array, meta=meta)
+
+            path = Path(tmpdir) / "bbox.mla"
+            image.save(path)
+
+            loaded = MLArray(path)
+            self.assertEqual(loaded.meta.bbox.to_plain(), meta.bbox.to_plain())
+
+
+if __name__ == "__main__":
+    unittest.main()