Add unit tests and improve docstrings for box_iou, box_giou, and box_pair_giou

Author: reworld223

monai/data/box_utils.py

@@ -826,7 +826,8 @@ def box_iou(boxes1: NdarrayOrTensor, boxes2: NdarrayOrTensor) -> NdarrayOrTensor
         boxes2: bounding boxes, Mx4 or Mx6 torch tensor or ndarray. The box mode is assumed to be ``StandardMode``
 
     Returns:
-        The output is always floating-point (size: (N, M)):
+        An array/tensor matching the container type of ``boxes1`` (NumPy ndarray or Torch tensor), always
+        floating-point with size ``(N, M)``:
         - if ``boxes1`` has a floating-point dtype, the same dtype is used.
         - if ``boxes1`` has an integer dtype, the result is returned as ``torch.float32``.
 
@@ -871,7 +872,8 @@ def box_giou(boxes1: NdarrayOrTensor, boxes2: NdarrayOrTensor) -> NdarrayOrTenso
         boxes2: bounding boxes, Mx4 or Mx6 torch tensor or ndarray. The box mode is assumed to be ``StandardMode``
 
     Returns:
-        The output is always floating-point (size: (N, M)):
+        An array/tensor matching the container type of ``boxes1`` (NumPy ndarray or Torch tensor), always
+        floating-point with size ``(N, M)``:
         - if ``boxes1`` has a floating-point dtype, the same dtype is used.
         - if ``boxes1`` has an integer dtype, the result is returned as ``torch.float32``.
 
@@ -934,7 +936,8 @@ def box_pair_giou(boxes1: NdarrayOrTensor, boxes2: NdarrayOrTensor) -> NdarrayOr
         boxes2: bounding boxes, same shape with boxes1. The box mode is assumed to be ``StandardMode``
 
     Returns:
-        The output is always floating-point (size: (N,)):
+        An array/tensor matching the container type of ``boxes1`` (NumPy ndarray or Torch tensor), always
+        floating-point with size ``(N, )``:
         - if ``boxes1`` has a floating-point dtype, the same dtype is used.
         - if ``boxes1`` has an integer dtype, the result is returned as ``torch.float32``.
 

tests/data/test_box_utils.py

@@ -14,6 +14,7 @@
 import unittest
 
 import numpy as np
+import torch
 from parameterized import parameterized
 
 from monai.data.box_utils import (
@@ -218,5 +219,46 @@ def test_value(self, input_data, mode2, expected_box, expected_area):
         assert_allclose(nms_box, [1], type_test=False)
 
 
+class TestBoxUtilsDtype(unittest.TestCase):
+    @parameterized.expand(
+        [
+            # numpy dtypes
+            (np.array([[0, 0, 0, 2, 2, 2]], dtype=np.int32), np.array([[0, 0, 0, 2, 2, 2]], dtype=np.int32)),
+            (np.array([[0, 0, 0, 2, 2, 2]], dtype=np.float32), np.array([[0, 0, 0, 2, 2, 2]], dtype=np.float32)),
+            # torch dtypes
+            (
+                torch.tensor([[0, 0, 0, 2, 2, 2]], dtype=torch.int64),
+                torch.tensor([[0, 0, 0, 2, 2, 2]], dtype=torch.int64),
+            ),
+            (
+                torch.tensor([[0, 0, 0, 2, 2, 2]], dtype=torch.float32),
+                torch.tensor([[0, 0, 0, 2, 2, 2]], dtype=torch.float32),
+            ),
+            # mixed numpy (int + float)
+            (np.array([[0, 0, 0, 2, 2, 2]], dtype=np.int32), np.array([[0, 0, 0, 2, 2, 2]], dtype=np.float32)),
+            # mixed torch (int + float)
+            (
+                torch.tensor([[0, 0, 0, 2, 2, 2]], dtype=torch.int64),
+                torch.tensor([[0, 0, 0, 2, 2, 2]], dtype=torch.float32),
+            ),
+        ]
+    )
+    def test_dtype_behavior(self, boxes1, boxes2):
+        funcs = [box_iou, box_giou, box_pair_giou]
+        for func in funcs:
+            result = func(boxes1, boxes2)
+
+            if isinstance(result, np.ndarray):
+                self.assertTrue(
+                    np.issubdtype(result.dtype, np.floating), f"{func.__name__} expected float, got {result.dtype}"
+                )
+            elif torch.is_tensor(result):
+                self.assertTrue(
+                    torch.is_floating_point(result), f"{func.__name__} expected float tensor, got {result.dtype}"
+                )
+            else:
+                self.fail(f"Unexpected return type {type(result)}")
+
+
 if __name__ == "__main__":
     unittest.main()