perf: shallow copy instead of deepcopy

perception_eval/perception_eval/common/deepcopy.py

@@ -0,0 +1,38 @@
+from copy import deepcopy
+
+
+# https://stackoverflow.com/a/24621200/4732868
+def deepcopy_with_sharing(obj, shared_attribute_names, memo=None):
+    """
+    Deepcopy an object, except for a given list of attributes, which should
+    be shared between the original object and its copy.
+
+    obj is some object
+    shared_attribute_names: A list of strings identifying the attributes that
+        should be shared between the original and its copy.
+    memo is the dictionary passed into __deepcopy__.  Ignore this argument if
+        not calling from within __deepcopy__.
+    """
+    assert isinstance(shared_attribute_names, (list, tuple))
+    shared_attributes = {k: getattr(obj, k) for k in shared_attribute_names}
+
+    if hasattr(obj, "__deepcopy__"):
+        # Do hack to prevent infinite recursion in call to deepcopy
+        deepcopy_method = obj.__deepcopy__
+        obj.__deepcopy__ = None
+
+    for attr in shared_attribute_names:
+        del obj.__dict__[attr]
+
+    clone = deepcopy(obj)
+
+    for attr, val in shared_attributes.items():
+        setattr(obj, attr, val)
+        setattr(clone, attr, val)
+
+    if hasattr(obj, "__deepcopy__"):
+        # Undo hack
+        obj.__deepcopy__ = deepcopy_method
+        del clone.__deepcopy__
+
+    return clone

perception_eval/perception_eval/evaluation/matching/objects_filter.py

@@ -12,6 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from copy import copy
 from typing import Dict
 from typing import List
 from typing import Optional
@@ -123,7 +124,7 @@ def filter_object_results(
             is_target = False
 
         if is_target:
-            filtered_object_results.append(object_result)
+            filtered_object_results.append(copy(object_result))
 
     return filtered_object_results
 
@@ -198,7 +199,7 @@ def filter_objects(
             transforms=transforms,
         )
         if is_target:
-            filtered_objects.append(object_)
+            filtered_objects.append(copy(object_))
     return filtered_objects
 
 

perception_eval/perception_eval/evaluation/result/object_result.py

@@ -25,6 +25,7 @@
 from perception_eval.common import DynamicObject
 from perception_eval.common import DynamicObject2D
 from perception_eval.common import ObjectType
+from perception_eval.common.deepcopy import deepcopy_with_sharing
 from perception_eval.common.evaluation_task import EvaluationTask
 from perception_eval.common.label import LabelType
 from perception_eval.common.label import TrafficLightLabel
@@ -106,6 +107,9 @@ def __init__(
             self.iou_3d = None
             self.plane_distance = None
 
+    def __deepcopy__(self, memo):
+        return deepcopy_with_sharing(self, shared_attribute_names = ['estimated_object', 'ground_truth_object'], memo=memo)
+
     def get_status(
         self,
         matching_mode: MatchingMode,

perception_eval/perception_eval/evaluation/result/perception_frame_result.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 from __future__ import annotations
-
+from copy import copy, deepcopy
 from typing import Dict
 from typing import List
 from typing import Optional

perception_eval/perception_eval/tool/utils.py

@@ -476,7 +476,6 @@ def filter_frame_by_distance(
         PerceptionFrameResult: Filtered frame results.
     """
     ret_frame = deepcopy(frame)
-
     if min_distance is not None:
         min_distance_list = [min_distance] * len(ret_frame.target_labels)
     else: