Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.8/site-packages/keras/src/layers/reshaping/up_sampling2d.py: 45%

1# Copyright 2015 The TensorFlow Authors. All Rights Reserved. 

2# 

3# Licensed under the Apache License, Version 2.0 (the "License"); 

4# you may not use this file except in compliance with the License. 

5# You may obtain a copy of the License at 

6# 

7# http://www.apache.org/licenses/LICENSE-2.0 

8# 

9# Unless required by applicable law or agreed to in writing, software 

10# distributed under the License is distributed on an "AS IS" BASIS, 

11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 

12# See the License for the specific language governing permissions and 

13# limitations under the License. 

14# ============================================================================== 

15"""Keras upsampling layer for 2D inputs.""" 

16 

17 

18import tensorflow.compat.v2 as tf 

19 

20from keras.src import backend 

21from keras.src.engine.base_layer import Layer 

22from keras.src.engine.input_spec import InputSpec 

23from keras.src.utils import conv_utils 

24from keras.src.utils import image_utils 

25 

26# isort: off 

27from tensorflow.python.util.tf_export import keras_export 

28 

29 

30@keras_export("keras.layers.UpSampling2D") 

31class UpSampling2D(Layer): 

32 """Upsampling layer for 2D inputs. 

33 

34 Repeats the rows and columns of the data 

35 by `size[0]` and `size[1]` respectively. 

36 

37 Examples: 

38 

39 >>> input_shape = (2, 2, 1, 3) 

40 >>> x = np.arange(np.prod(input_shape)).reshape(input_shape) 

41 >>> print(x) 

42 [[[[ 0 1 2]] 

43 [[ 3 4 5]]] 

44 [[[ 6 7 8]] 

45 [[ 9 10 11]]]] 

46 >>> y = tf.keras.layers.UpSampling2D(size=(1, 2))(x) 

47 >>> print(y) 

48 tf.Tensor( 

49 [[[[ 0 1 2] 

50 [ 0 1 2]] 

51 [[ 3 4 5] 

52 [ 3 4 5]]] 

53 [[[ 6 7 8] 

54 [ 6 7 8]] 

55 [[ 9 10 11] 

56 [ 9 10 11]]]], shape=(2, 2, 2, 3), dtype=int64) 

57 

58 Args: 

59 size: Int, or tuple of 2 integers. 

60 The upsampling factors for rows and columns. 

61 data_format: A string, 

62 one of `channels_last` (default) or `channels_first`. 

63 The ordering of the dimensions in the inputs. 

64 `channels_last` corresponds to inputs with shape 

65 `(batch_size, height, width, channels)` while `channels_first` 

66 corresponds to inputs with shape 

67 `(batch_size, channels, height, width)`. 

68 When unspecified, uses 

69 `image_data_format` value found in your Keras config file at 

70 `~/.keras/keras.json` (if exists) else 'channels_last'. 

71 Defaults to 'channels_last'. 

72 interpolation: A string, one of `"area"`, `"bicubic"`, `"bilinear"`, 

73 `"gaussian"`, `"lanczos3"`, `"lanczos5"`, `"mitchellcubic"`, 

74 `"nearest"`. 

75 

76 Input shape: 

77 4D tensor with shape: 

78 - If `data_format` is `"channels_last"`: 

79 `(batch_size, rows, cols, channels)` 

80 - If `data_format` is `"channels_first"`: 

81 `(batch_size, channels, rows, cols)` 

82 

83 Output shape: 

84 4D tensor with shape: 

85 - If `data_format` is `"channels_last"`: 

86 `(batch_size, upsampled_rows, upsampled_cols, channels)` 

87 - If `data_format` is `"channels_first"`: 

88 `(batch_size, channels, upsampled_rows, upsampled_cols)` 

89 """ 

90 

91 def __init__( 

92 self, size=(2, 2), data_format=None, interpolation="nearest", **kwargs 

93 ): 

94 super().__init__(**kwargs) 

95 self.data_format = conv_utils.normalize_data_format(data_format) 

96 self.size = conv_utils.normalize_tuple(size, 2, "size") 

97 self.interpolation = image_utils.get_interpolation(interpolation) 

98 self.input_spec = InputSpec(ndim=4) 

99 

100 def compute_output_shape(self, input_shape): 

101 input_shape = tf.TensorShape(input_shape).as_list() 

102 if self.data_format == "channels_first": 

103 height = ( 

104 self.size[0] * input_shape[2] 

105 if input_shape[2] is not None 

106 else None 

107 ) 

108 width = ( 

109 self.size[1] * input_shape[3] 

110 if input_shape[3] is not None 

111 else None 

112 ) 

113 return tf.TensorShape( 

114 [input_shape[0], input_shape[1], height, width] 

115 ) 

116 else: 

117 height = ( 

118 self.size[0] * input_shape[1] 

119 if input_shape[1] is not None 

120 else None 

121 ) 

122 width = ( 

123 self.size[1] * input_shape[2] 

124 if input_shape[2] is not None 

125 else None 

126 ) 

127 return tf.TensorShape( 

128 [input_shape[0], height, width, input_shape[3]] 

129 ) 

130 

131 def call(self, inputs): 

132 return backend.resize_images( 

133 inputs, 

134 self.size[0], 

135 self.size[1], 

136 self.data_format, 

137 interpolation=self.interpolation, 

138 ) 

139 

140 def get_config(self): 

141 config = { 

142 "size": self.size, 

143 "data_format": self.data_format, 

144 "interpolation": self.interpolation, 

145 } 

146 base_config = super().get_config() 

147 return dict(list(base_config.items()) + list(config.items())) 

148