diff --git a/test/test_transforms_v2.py b/test/test_transforms_v2.py
index bca8925c71d..b5a8c9ead7e 100644
--- a/test/test_transforms_v2.py
+++ b/test/test_transforms_v2.py
@@ -2021,6 +2021,9 @@ def test_sanitize_bounding_boxes(min_size, labels_getter, sample_type):
     assert out_image is input_img
     assert out_whatever is whatever
 
+    assert isinstance(out_boxes, datapoints.BoundingBox)
+    assert isinstance(out_masks, datapoints.Mask)
+
     if labels_getter is None or (callable(labels_getter) and labels_getter({"labels": "blah"}) is None):
         assert out_labels is labels
     else:
diff --git a/torchvision/transforms/v2/_misc.py b/torchvision/transforms/v2/_misc.py
index c9b9025ebd9..8f063fd6049 100644
--- a/torchvision/transforms/v2/_misc.py
+++ b/torchvision/transforms/v2/_misc.py
@@ -397,10 +397,15 @@ def forward(self, *inputs: Any) -> Any:
         return tree_unflatten(flat_outputs, spec)
 
     def _transform(self, inpt: Any, params: Dict[str, Any]) -> Any:
+        is_label = inpt is not None and inpt is params["labels"]
+        is_bounding_box_or_mask = isinstance(inpt, (datapoints.BoundingBox, datapoints.Mask))
 
-        if (inpt is not None and inpt is params["labels"]) or isinstance(
-            inpt, (datapoints.BoundingBox, datapoints.Mask)
-        ):
-            inpt = inpt[params["valid"]]
+        if not (is_label or is_bounding_box_or_mask):
+            return inpt
 
-        return inpt
+        output = inpt[params["valid"]]
+
+        if is_label:
+            return output
+
+        return type(inpt).wrap_like(inpt, output)