Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.8/site-packages/keras/src/applications/vgg16.py: 28%
72 statements
« prev ^ index » next coverage.py v7.4.0, created at 2024-01-03 07:57 +0000
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# ==============================================================================
16"""VGG16 model for Keras.
18Reference:
19 - [Very Deep Convolutional Networks for Large-Scale Image Recognition]
20 (https://arxiv.org/abs/1409.1556) (ICLR 2015)
21"""
23import tensorflow.compat.v2 as tf
25from keras.src import backend
26from keras.src.applications import imagenet_utils
27from keras.src.engine import training
28from keras.src.layers import VersionAwareLayers
29from keras.src.utils import data_utils
30from keras.src.utils import layer_utils
32# isort: off
33from tensorflow.python.util.tf_export import keras_export
35WEIGHTS_PATH = (
36 "https://storage.googleapis.com/tensorflow/keras-applications/"
37 "vgg16/vgg16_weights_tf_dim_ordering_tf_kernels.h5"
38)
39WEIGHTS_PATH_NO_TOP = (
40 "https://storage.googleapis.com/tensorflow/"
41 "keras-applications/vgg16/"
42 "vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5"
43)
45layers = VersionAwareLayers()
48@keras_export("keras.applications.vgg16.VGG16", "keras.applications.VGG16")
49def VGG16(
50 include_top=True,
51 weights="imagenet",
52 input_tensor=None,
53 input_shape=None,
54 pooling=None,
55 classes=1000,
56 classifier_activation="softmax",
57):
58 """Instantiates the VGG16 model.
60 Reference:
61 - [Very Deep Convolutional Networks for Large-Scale Image Recognition](
62 https://arxiv.org/abs/1409.1556) (ICLR 2015)
64 For image classification use cases, see
65 [this page for detailed examples](
66 https://keras.io/api/applications/#usage-examples-for-image-classification-models).
68 For transfer learning use cases, make sure to read the
69 [guide to transfer learning & fine-tuning](
70 https://keras.io/guides/transfer_learning/).
72 The default input size for this model is 224x224.
74 Note: each Keras Application expects a specific kind of input preprocessing.
75 For VGG16, call `tf.keras.applications.vgg16.preprocess_input` on your
76 inputs before passing them to the model.
77 `vgg16.preprocess_input` will convert the input images from RGB to BGR,
78 then will zero-center each color channel with respect to the ImageNet
79 dataset, without scaling.
81 Args:
82 include_top: whether to include the 3 fully-connected
83 layers at the top of the network.
84 weights: one of `None` (random initialization),
85 'imagenet' (pre-training on ImageNet),
86 or the path to the weights file to be loaded.
87 input_tensor: optional Keras tensor
88 (i.e. output of `layers.Input()`)
89 to use as image input for the model.
90 input_shape: optional shape tuple, only to be specified
91 if `include_top` is False (otherwise the input shape
92 has to be `(224, 224, 3)`
93 (with `channels_last` data format)
94 or `(3, 224, 224)` (with `channels_first` data format).
95 It should have exactly 3 input channels,
96 and width and height should be no smaller than 32.
97 E.g. `(200, 200, 3)` would be one valid value.
98 pooling: Optional pooling mode for feature extraction
99 when `include_top` is `False`.
100 - `None` means that the output of the model will be
101 the 4D tensor output of the
102 last convolutional block.
103 - `avg` means that global average pooling
104 will be applied to the output of the
105 last convolutional block, and thus
106 the output of the model will be a 2D tensor.
107 - `max` means that global max pooling will
108 be applied.
109 classes: optional number of classes to classify images
110 into, only to be specified if `include_top` is True, and
111 if no `weights` argument is specified.
112 classifier_activation: A `str` or callable. The activation function to
113 use on the "top" layer. Ignored unless `include_top=True`. Set
114 `classifier_activation=None` to return the logits of the "top"
115 layer. When loading pretrained weights, `classifier_activation` can
116 only be `None` or `"softmax"`.
118 Returns:
119 A `keras.Model` instance.
120 """
121 if not (weights in {"imagenet", None} or tf.io.gfile.exists(weights)):
122 raise ValueError(
123 "The `weights` argument should be either "
124 "`None` (random initialization), `imagenet` "
125 "(pre-training on ImageNet), "
126 "or the path to the weights file to be loaded. Received: "
127 f"weights={weights}"
128 )
130 if weights == "imagenet" and include_top and classes != 1000:
131 raise ValueError(
132 'If using `weights` as `"imagenet"` with `include_top` '
133 "as true, `classes` should be 1000. "
134 f"Received `classes={classes}`"
135 )
136 # Determine proper input shape
137 input_shape = imagenet_utils.obtain_input_shape(
138 input_shape,
139 default_size=224,
140 min_size=32,
141 data_format=backend.image_data_format(),
142 require_flatten=include_top,
143 weights=weights,
144 )
146 if input_tensor is None:
147 img_input = layers.Input(shape=input_shape)
148 else:
149 if not backend.is_keras_tensor(input_tensor):
150 img_input = layers.Input(tensor=input_tensor, shape=input_shape)
151 else:
152 img_input = input_tensor
153 # Block 1
154 x = layers.Conv2D(
155 64, (3, 3), activation="relu", padding="same", name="block1_conv1"
156 )(img_input)
157 x = layers.Conv2D(
158 64, (3, 3), activation="relu", padding="same", name="block1_conv2"
159 )(x)
160 x = layers.MaxPooling2D((2, 2), strides=(2, 2), name="block1_pool")(x)
162 # Block 2
163 x = layers.Conv2D(
164 128, (3, 3), activation="relu", padding="same", name="block2_conv1"
165 )(x)
166 x = layers.Conv2D(
167 128, (3, 3), activation="relu", padding="same", name="block2_conv2"
168 )(x)
169 x = layers.MaxPooling2D((2, 2), strides=(2, 2), name="block2_pool")(x)
171 # Block 3
172 x = layers.Conv2D(
173 256, (3, 3), activation="relu", padding="same", name="block3_conv1"
174 )(x)
175 x = layers.Conv2D(
176 256, (3, 3), activation="relu", padding="same", name="block3_conv2"
177 )(x)
178 x = layers.Conv2D(
179 256, (3, 3), activation="relu", padding="same", name="block3_conv3"
180 )(x)
181 x = layers.MaxPooling2D((2, 2), strides=(2, 2), name="block3_pool")(x)
183 # Block 4
184 x = layers.Conv2D(
185 512, (3, 3), activation="relu", padding="same", name="block4_conv1"
186 )(x)
187 x = layers.Conv2D(
188 512, (3, 3), activation="relu", padding="same", name="block4_conv2"
189 )(x)
190 x = layers.Conv2D(
191 512, (3, 3), activation="relu", padding="same", name="block4_conv3"
192 )(x)
193 x = layers.MaxPooling2D((2, 2), strides=(2, 2), name="block4_pool")(x)
195 # Block 5
196 x = layers.Conv2D(
197 512, (3, 3), activation="relu", padding="same", name="block5_conv1"
198 )(x)
199 x = layers.Conv2D(
200 512, (3, 3), activation="relu", padding="same", name="block5_conv2"
201 )(x)
202 x = layers.Conv2D(
203 512, (3, 3), activation="relu", padding="same", name="block5_conv3"
204 )(x)
205 x = layers.MaxPooling2D((2, 2), strides=(2, 2), name="block5_pool")(x)
207 if include_top:
208 # Classification block
209 x = layers.Flatten(name="flatten")(x)
210 x = layers.Dense(4096, activation="relu", name="fc1")(x)
211 x = layers.Dense(4096, activation="relu", name="fc2")(x)
213 imagenet_utils.validate_activation(classifier_activation, weights)
214 x = layers.Dense(
215 classes, activation=classifier_activation, name="predictions"
216 )(x)
217 else:
218 if pooling == "avg":
219 x = layers.GlobalAveragePooling2D()(x)
220 elif pooling == "max":
221 x = layers.GlobalMaxPooling2D()(x)
223 # Ensure that the model takes into account
224 # any potential predecessors of `input_tensor`.
225 if input_tensor is not None:
226 inputs = layer_utils.get_source_inputs(input_tensor)
227 else:
228 inputs = img_input
229 # Create model.
230 model = training.Model(inputs, x, name="vgg16")
232 # Load weights.
233 if weights == "imagenet":
234 if include_top:
235 weights_path = data_utils.get_file(
236 "vgg16_weights_tf_dim_ordering_tf_kernels.h5",
237 WEIGHTS_PATH,
238 cache_subdir="models",
239 file_hash="64373286793e3c8b2b4e3219cbf3544b",
240 )
241 else:
242 weights_path = data_utils.get_file(
243 "vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5",
244 WEIGHTS_PATH_NO_TOP,
245 cache_subdir="models",
246 file_hash="6d6bbae143d832006294945121d1f1fc",
247 )
248 model.load_weights(weights_path)
249 elif weights is not None:
250 model.load_weights(weights)
252 return model
255@keras_export("keras.applications.vgg16.preprocess_input")
256def preprocess_input(x, data_format=None):
257 return imagenet_utils.preprocess_input(
258 x, data_format=data_format, mode="caffe"
259 )
262@keras_export("keras.applications.vgg16.decode_predictions")
263def decode_predictions(preds, top=5):
264 return imagenet_utils.decode_predictions(preds, top=top)
267preprocess_input.__doc__ = imagenet_utils.PREPROCESS_INPUT_DOC.format(
268 mode="",
269 ret=imagenet_utils.PREPROCESS_INPUT_RET_DOC_CAFFE,
270 error=imagenet_utils.PREPROCESS_INPUT_ERROR_DOC,
271)
272decode_predictions.__doc__ = imagenet_utils.decode_predictions.__doc__