Merge pull request #4635 from thewtex/transform-dict-repr

BUG: Make dict_from_transform more consistent with other dict representations

Author: thewtex

Modules/Core/Transform/wrapping/test/itkTransformSerializationTest.py

@@ -35,16 +35,13 @@
 
 keys_to_test1 = [
     "name",
-    "parametersValueType",
-    "transformType",
-    "inputDimension",
-    "outputDimension",
     "inputSpaceName",
     "outputSpaceName",
     "numberOfParameters",
     "numberOfFixedParameters",
 ]
 keys_to_test2 = ["parameters", "fixedParameters"]
+keys_to_test3 = ["transformParameterization", "parametersValueType", "inputDimension", "outputDimension"]
 
 transform_object_list = []
 for i, transform_type in enumerate(transforms_to_test):
@@ -60,6 +57,8 @@
     # Test all the parameters
     for k in keys_to_test2:
         assert np.array_equal(serialize_deserialize[k], transform[k])
+    for k in keys_to_test3:
+        assert serialize_deserialize["transformType"][k], transform["transformType"][k]
     transform_object_list.append(transform)
 
 print("Individual Transforms Test Done")
@@ -93,6 +92,9 @@
     for k in keys_to_test2:
         assert np.array_equal(transform_obj[k], transform_object_list[i][k])
 
+    for k in keys_to_test3:
+        assert transform_object_list[i]["transformType"][k], transform["transformType"][k]
+
 
 # Test for transformation using de-serialized BSpline Transform
 ImageDimension = 2

Wrapping/Generators/Python/PyBase/pyBase.i

@@ -430,15 +430,15 @@ str = str
                 Return keys related to the transform's metadata.
                 These keys are used in the dictionary resulting from dict(transform).
                 """
-                result = ['name', 'inputDimension', 'outputDimension', 'inputSpaceName', 'outputSpaceName', 'numberOfParameters', 'numberOfFixedParameters', 'parameters', 'fixedParameters']
+                result = ['transformType', 'name', 'inputSpaceName', 'outputSpaceName', 'numberOfParameters', 'numberOfFixedParameters', 'parameters', 'fixedParameters']
                 return result
 
             def __getitem__(self, key):
                 """Access metadata keys, see help(transform.keys), for string keys."""
                 import itk
                 if isinstance(key, str):
                     state = itk.dict_from_transform(self)
-                    return state[0][key]
+                    return state[key]
 
             def __setitem__(self, key, value):
                 if isinstance(key, str):
@@ -474,7 +474,6 @@ str = str
             def __setstate__(self, state):
                 """Set object state, necessary for serialization with pickle."""
                 import itk
-                import numpy as np
                 deserialized = itk.transform_from_dict(state)
                 self.__dict__['this'] = deserialized
             %}

Wrapping/Generators/Python/Tests/extras.py

@@ -343,6 +343,11 @@ def custom_callback(name, progress):
 parameters = np.asarray(transforms[0].GetParameters())
 assert np.allclose(parameters, np.array(baseline_additional_transform_params))
 
+transform_dict = itk.dict_from_transform(transforms[0])
+transform_back = itk.transform_from_dict(transform_dict)
+transform_dict = itk.dict_from_transform(transforms)
+transform_back = itk.transform_from_dict(transform_dict)
+
 # pipeline, auto_pipeline and templated class are tested in other files
 
 # BridgeNumPy

Wrapping/Generators/Python/itk/support/extras.py

@@ -981,57 +981,81 @@ def dict_from_pointset(pointset: "itkt.PointSet") -> Dict:
     )
 
 
-def dict_from_transform(transform: "itkt.TransformBase") -> Dict:
+def dict_from_transform(transform: Union["itkt.TransformBase", List["itkt.TransformBase"]]) -> Union[List[Dict], Dict]:
+    """Serialize a Python itk.Transform object to a pickable Python dictionary.
+
+    If the transform is a list of transforms, then a list of dictionaries is returned.
+    If the transform is a single, non-Composite transform, then a single dictionary is returned.
+    Composite transforms and nested composite transforms are flattened into a list of dictionaries.
+    """
     import itk
+    datatype_dict = {"double": itk.D, "float": itk.F}
 
     def update_transform_dict(current_transform):
         current_transform_type = current_transform.GetTransformTypeAsString()
         current_transform_type_split = current_transform_type.split("_")
-        component = itk.template(current_transform)
 
-        in_transform_dict = dict()
-        in_transform_dict["name"] = current_transform.GetObjectName()
+        transform_type = dict()
+        transform_parameterization = current_transform_type_split[0].replace("Transform", "")
+        transform_type["transformParameterization"] = transform_parameterization
 
-        datatype_dict = {"double": itk.D, "float": itk.F}
-        in_transform_dict["parametersValueType"] = python_to_js(
+        transform_type["parametersValueType"] = python_to_js(
             datatype_dict[current_transform_type_split[1]]
         )
-        in_transform_dict["inputDimension"] = int(current_transform_type_split[2])
-        in_transform_dict["outputDimension"] = int(current_transform_type_split[3])
-        in_transform_dict["transformType"] = current_transform_type_split[0]
+        transform_type["inputDimension"] = int(current_transform_type_split[2])
+        transform_type["outputDimension"] = int(current_transform_type_split[3])
 
-        in_transform_dict["inputSpaceName"] = current_transform.GetInputSpaceName()
-        in_transform_dict["outputSpaceName"] = current_transform.GetOutputSpaceName()
+        transform_dict = dict()
+        transform_dict['transformType'] = transform_type
+        transform_dict["name"] = current_transform.GetObjectName()
+
+        transform_dict["inputSpaceName"] = current_transform.GetInputSpaceName()
+        transform_dict["outputSpaceName"] = current_transform.GetOutputSpaceName()
 
         # To avoid copying the parameters for the Composite Transform
         # as it is a copy of child transforms.
         if "Composite" not in current_transform_type_split[0]:
             p = np.array(current_transform.GetParameters())
-            in_transform_dict["parameters"] = p
+            transform_dict["parameters"] = p
 
             fp = np.array(current_transform.GetFixedParameters())
-            in_transform_dict["fixedParameters"] = fp
+            transform_dict["fixedParameters"] = fp
 
-            in_transform_dict["numberOfParameters"] = p.shape[0]
-            in_transform_dict["numberOfFixedParameters"] = fp.shape[0]
+            transform_dict["numberOfParameters"] = p.shape[0]
+            transform_dict["numberOfFixedParameters"] = fp.shape[0]
 
-        return in_transform_dict
+        return transform_dict
 
     dict_array = []
-    transform_type = transform.GetTransformTypeAsString()
-    if "CompositeTransform" in transform_type:
-        # Add the transforms inside the composite transform
-        # range is over-ridden so using this hack to create a list
-        for i, _ in enumerate([0] * transform.GetNumberOfTransforms()):
-            current_transform = transform.GetNthTransform(i)
-            dict_array.append(update_transform_dict(current_transform))
+    multi = False
+    def add_transform_dict(transform):
+        transform_type = transform.GetTransformTypeAsString()
+        if "CompositeTransform" in transform_type:
+            # Add the transforms inside the composite transform
+            # range is over-ridden so using this hack to create a list
+            for i, _ in enumerate([0] * transform.GetNumberOfTransforms()):
+                current_transform = transform.GetNthTransform(i)
+                dict_array.append(update_transform_dict(current_transform))
+            return True
+        else:
+            dict_array.append(update_transform_dict(transform))
+            return False
+    if isinstance(transform, list):
+        multi = True
+        for t in transform:
+            add_transform_dict(t)
     else:
-        dict_array.append(update_transform_dict(transform))
+        multi = add_transform_dict(transform)
 
-    return dict_array
+    if multi:
+        return dict_array
+    else:
+        return dict_array[0]
 
+def transform_from_dict(transform_dict: Union[Dict, List[Dict]]) -> "itkt.TransformBase":
+    """Deserialize a dictionary representing an itk.Transform object.
 
-def transform_from_dict(transform_dict: Dict) -> "itkt.TransformBase":
+    If the dictionary represents a list of transforms, then a Composite Transform is returned."""
     import itk
 
     def set_parameters(transform, transform_parameters, transform_fixed_parameters, data_type):
@@ -1055,35 +1079,41 @@ def special_transform_check(transform_name):
 
     parametersValueType_dict = {"float32": itk.F, "float64": itk.D}
 
+    if not isinstance(transform_dict, list):
+        transform_dict = [transform_dict]
+
     # Loop over all the transforms in the dictionary
     transforms_list = []
     for i, _ in enumerate(transform_dict):
-        data_type = parametersValueType_dict[transform_dict[i]["parametersValueType"]]
+        transform_type = transform_dict[i]["transformType"]
+        data_type = parametersValueType_dict[transform_type["parametersValueType"]]
+
+        transform_parameterization = transform_type["transformParameterization"] + 'Transform'
 
         # No template parameter needed for transforms having 2D or 3D name
         # Also for some selected transforms
-        if special_transform_check(transform_dict[i]["transformType"]):
-            transform_template = getattr(itk, transform_dict[i]["transformType"])
+        if special_transform_check(transform_parameterization):
+            transform_template = getattr(itk, transform_parameterization)
             transform = transform_template[data_type].New()
         # Currently only BSpline Transform has 3 template parameters
         # For future extensions the information will have to be encoded in
         # the transformType variable. The transform object once added in a
         # composite transform lose the information for other template parameters ex. BSpline.
         # The Spline order is fixed as 3 here.
-        elif transform_dict[i]["transformType"] == "BSplineTransform":
-            transform_template = getattr(itk, transform_dict[i]["transformType"])
+        elif transform_parameterization == "BSplineTransform":
+            transform_template = getattr(itk, transform_parameterization)
             transform = transform_template[
-                data_type, transform_dict[i]["inputDimension"], 3
+                data_type, transform_type["inputDimension"], 3
             ].New()
         else:
-            transform_template = getattr(itk, transform_dict[i]["transformType"])
+            transform_template = getattr(itk, transform_parameterization)
             if len(transform_template.items()[0][0]) > 2:
                 transform = transform_template[
-                    data_type, transform_dict[i]["inputDimension"], transform_dict[i]["outputDimension"]
+                    data_type, transform_type["inputDimension"], transform_type["outputDimension"]
                 ].New()
             else:
                 transform = transform_template[
-                    data_type, transform_dict[i]["inputDimension"]
+                    data_type, transform_type["inputDimension"]
                 ].New()
 
         transform.SetObjectName(transform_dict[i]["name"])
@@ -1102,8 +1132,8 @@ def special_transform_check(transform_name):
     if len(transforms_list) > 1:
         # Create a Composite Transform object
         # and add all the transforms in it.
-        data_type = parametersValueType_dict[transform_dict[0]["parametersValueType"]]
-        transform = itk.CompositeTransform[data_type, transforms_list[0]['inputDimension']].New()
+        data_type = parametersValueType_dict[transform_dict[0]["transformType"]["parametersValueType"]]
+        transform = itk.CompositeTransform[data_type, transforms_list[0]["transformType"]['inputDimension']].New()
         for current_transform in transforms_list:
             transform.AddTransform(current_transform)
     else: