better typehints for numpy arrays in Sample classes

upgrade datumaro

Author: leoll2

application/backend/app/services/datumaro_converter.py

@@ -21,6 +21,7 @@
 
 from app.models import DatasetItem, DatasetItemSubset, Label, Polygon, Rectangle, TaskType
 from app.schemas.project import TaskBase
+from app.utils.typing import NDArrayFloat32, NDArrayInt
 
 CONVERSION_BATCH_SIZE = 50
 
@@ -56,9 +57,8 @@ class MultilabelClassificationSample(Sample):
 
     image: str = image_path_field()
     image_info: ImageInfo = image_info_field()
-    # TODO: Use NDArrayFloat32 and NDArrayInt instead of np.ndarray after open-edge-platform/datumaro#1949 is solved
-    label: np.ndarray = label_field(dtype=pl.Int32(), multi_label=True)
-    confidence: np.ndarray | None = score_field(dtype=pl.Float32(), is_list=True)
+    label: NDArrayInt = label_field(dtype=pl.Int32(), multi_label=True)
+    confidence: NDArrayFloat32 | None = score_field(dtype=pl.Float32(), is_list=True)
     subset: Subset = subset_field()
 
 
@@ -76,10 +76,9 @@ class DetectionSample(Sample):
 
     image: str = image_path_field()
     image_info: ImageInfo = image_info_field()
-    # TODO: Use NDArrayFloat32 and NDArrayInt instead of np.ndarray after open-edge-platform/datumaro#1949 is solved
-    bboxes: np.ndarray = bbox_field(dtype=pl.Int32())
-    label: np.ndarray = label_field(dtype=pl.Int32(), is_list=True)
-    confidence: np.ndarray | None = score_field(dtype=pl.Float32(), is_list=True)
+    bboxes: NDArrayInt = bbox_field(dtype=pl.Int32())
+    label: NDArrayInt = label_field(dtype=pl.Int32(), is_list=True)
+    confidence: NDArrayFloat32 | None = score_field(dtype=pl.Float32(), is_list=True)
     subset: Subset = subset_field()
 
 
@@ -97,10 +96,9 @@ class InstanceSegmentationSample(Sample):
 
     image: str = image_path_field()
     image_info: ImageInfo = image_info_field()
-    # TODO: Use NDArrayFloat32 and NDArrayInt instead of np.ndarray after open-edge-platform/datumaro#1949 is solved
-    polygons: np.ndarray = polygon_field(dtype=pl.Float32())
-    label: np.ndarray = label_field(dtype=pl.Int32(), is_list=True)
-    confidence: np.ndarray | None = score_field(dtype=pl.Float32(), is_list=True)
+    polygons: NDArrayFloat32 = polygon_field(dtype=pl.Float32())
+    label: NDArrayInt = label_field(dtype=pl.Int32(), is_list=True)
+    confidence: NDArrayFloat32 | None = score_field(dtype=pl.Float32(), is_list=True)
     subset: Subset = subset_field()