@@ -71,6 +71,8 @@ def __call__(self, input_dict, label_dict, weight_dict):
7171class Normalize :
7272 """Normalize data class.
7373
74+ NOTE: This transform will modify the input data dict inplace.
75+
7476 Args:
7577 mean (Union[np.ndarray, Tuple[float, ...]]): Mean of training dataset.
7678 std (Union[np.ndarray, Tuple[float, ...]]): Standard Deviation of training dataset.
@@ -96,20 +98,20 @@ def __init__(
9698 self .apply_keys = apply_keys
9799
98100 def __call__ (self , input_item , label_item , weight_item ):
99- input_item_copy = {** input_item }
100- label_item_copy = {** label_item }
101101 if "input" in self .apply_keys :
102- for key , value in input_item_copy .items ():
103- input_item_copy [key ] = (value - self .mean ) / self .std
102+ for key , value in input_item .items ():
103+ input_item [key ] = (value - self .mean ) / self .std
104104 if "label" in self .apply_keys :
105- for key , value in label_item_copy .items ():
106- label_item_copy [key ] = (value - self .mean ) / self .std
107- return input_item_copy , label_item_copy , weight_item
105+ for key , value in label_item .items ():
106+ label_item [key ] = (value - self .mean ) / self .std
107+ return input_item , label_item , weight_item
108108
109109
110110class Log1p :
111111 """Calculates the natural logarithm of one plus the data, element-wise.
112112
113+ NOTE: This transform will modify the input data dict inplace.
114+
113115 Args:
114116 scale (float, optional): Scale data. Defaults to 1.0.
115117 apply_keys (Tuple[str, ...], optional): Which data is the log1p method applied to. Defaults to ("input", "label").
@@ -132,22 +134,22 @@ def __init__(
132134 self .apply_keys = apply_keys
133135
134136 def __call__ (self , input_item , label_item , weight_item ):
135- input_item_copy = {** input_item }
136- label_item_copy = {** label_item }
137137 if "input" in self .apply_keys :
138- for key , value in input_item_copy .items ():
139- input_item_copy [key ] = np .log1p (value / self .scale )
138+ for key , value in input_item .items ():
139+ input_item [key ] = np .log1p (value / self .scale )
140140 if "label" in self .apply_keys :
141- for key , value in label_item_copy .items ():
142- label_item_copy [key ] = np .log1p (value / self .scale )
143- return input_item_copy , label_item_copy , weight_item
141+ for key , value in label_item .items ():
142+ label_item [key ] = np .log1p (value / self .scale )
143+ return input_item , label_item , weight_item
144144
145145
146146class CropData :
147147 """Crop data class.
148148
149149 This class is used to crop data based on a specified bounding box.
150150
151+ NOTE: This transform will modify the input data dict inplace.
152+
151153 Args:
152154 xmin (Tuple[int, ...]): Bottom left corner point, [x0, y0].
153155 xmax (Tuple[int, ...]): Top right corner point, [x1, y1].
@@ -182,24 +184,24 @@ def __init__(
182184 self .apply_keys = apply_keys
183185
184186 def __call__ (self , input_item , label_item , weight_item ):
185- input_item_copy = {** input_item }
186- label_item_copy = {** label_item }
187187 if "input" in self .apply_keys :
188- for key , value in input_item_copy .items ():
189- input_item_copy [key ] = value [
188+ for key , value in input_item .items ():
189+ input_item [key ] = value [
190190 :, self .xmin [0 ] : self .xmax [0 ], self .xmin [1 ] : self .xmax [1 ]
191191 ]
192192 if "label" in self .apply_keys :
193- for key , value in label_item_copy .items ():
194- label_item_copy [key ] = value [
193+ for key , value in label_item .items ():
194+ label_item [key ] = value [
195195 :, self .xmin [0 ] : self .xmax [0 ], self .xmin [1 ] : self .xmax [1 ]
196196 ]
197- return input_item_copy , label_item_copy , weight_item
197+ return input_item , label_item , weight_item
198198
199199
200200class SqueezeData :
201201 """Squeeze data class.
202202
203+ NOTE: This transform will modify the input data dict inplace.
204+
203205 Args:
204206 apply_keys (Tuple[str, ...], optional): Which data is the squeeze method applied to. Defaults to ("input", "label").
205207
@@ -216,27 +218,25 @@ def __init__(self, apply_keys: Tuple[str, ...] = ("input", "label")):
216218 self .apply_keys = apply_keys
217219
218220 def __call__ (self , input_item , label_item , weight_item ):
219- input_item_copy = {** input_item }
220- label_item_copy = {** label_item }
221221 if "input" in self .apply_keys :
222- for key , value in input_item_copy .items ():
222+ for key , value in input_item .items ():
223223 if value .ndim == 4 :
224224 B , C , H , W = value .shape
225- input_item_copy [key ] = value .reshape ((B * C , H , W ))
225+ input_item [key ] = value .reshape ((B * C , H , W ))
226226 if value .ndim != 3 :
227227 raise ValueError (
228228 f"Only support squeeze data to ndim=3 now, but got ndim={ value .ndim }
229229 )
230230 if "label" in self .apply_keys :
231- for key , value in label_item_copy .items ():
231+ for key , value in label_item .items ():
232232 if value .ndim == 4 :
233233 B , C , H , W = value .shape
234- label_item_copy [key ] = value .reshape ((B * C , H , W ))
234+ label_item [key ] = value .reshape ((B * C , H , W ))
235235 if value .ndim != 3 :
236236 raise ValueError (
237237 f"Only support squeeze data to ndim=3 now, but got ndim={ value .ndim }
238238 )
239- return input_item_copy , label_item_copy , weight_item
239+ return input_item , label_item , weight_item
240240
241241
242242class FunctionalTransform :
0 commit comments