def squeeze_net(nb_classes, input_shape, dropout, input_tensor=None):
model_input = Input(input_shape) if input_tensor is None else input_tensor
x = BatchNormalization()(model_input)
x = Conv2D(64, kernel_size=(3, 3), padding='same', activation='relu')(x)
x = MaxPool2D()(x)
x = fire_module(x, (16, 64, 64), dropout)
x = fire_module(x, (16, 64, 64), dropout)
x = MaxPool2D(pool_size=(2, 2))(x)
x = fire_module(x, (32, 128, 128), dropout)
x = fire_module(x, (32, 128, 128), dropout)
x = MaxPool2D(pool_size=(2, 2))(x)
x = fire_module(x, (48, 192, 192), dropout)
x = fire_module(x, (48, 192, 192), dropout)
x = fire_module(x, (64, 256, 256), dropout)
x = fire_module(x, (64, 256, 256), dropout)
x = Conv2D(192, kernel_size=(1, 1), activation='relu')(x)
x = GlobalAveragePooling2D()(x)
x = Dense(32, activation='relu')(x)
x = Dense(nb_classes, activation='sigmoid')(x)
squeezenet = models.Model(get_source_inputs(model_input), x)
return squeezenet
def PReLUDepthNet(input_tensor=None,
input_shape=None,
width=4,
depth=50,
classes=1000):
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
x = Conv2D(width, (7, 7), strides=(2, 2), padding='same',
name='convfirst')(img_input)
x = PReLU()(x)
for i in range(depth):
x = Conv2D(width, (3, 3), padding='same', name='conv{}'.format(i))(x)
x = PReLU()(x)
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='fc')(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='preludepthnet-{}-{}'.format(width, depth))
return model
def nvidia_net(nb_classes,
filter_multiplicator,
input_shape,
dropout,
input_tensor=None):
model_input = Input(input_shape) if input_tensor is None else input_tensor
x = Conv2D(int(np.max([24 * filter_multiplicator, 1])),
kernel_size=(5, 5),
padding='same')(model_input)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = MaxPool2D()(x)
x = Conv2D(int(np.max([36 * filter_multiplicator, 1])),
kernel_size=(5, 5),
padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = MaxPool2D()(x)
x = Conv2D(int(np.max([48 * filter_multiplicator, 1])),
kernel_size=(5, 5),
padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = MaxPool2D()(x)
x = Conv2D(int(np.max([64 * filter_multiplicator, 1])),
kernel_size=(3, 3),
padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = MaxPool2D()(x)
x = Conv2D(int(np.max([64 * filter_multiplicator, 1])),
kernel_size=(3, 3),
padding='same')(x)
x = BatchNormalization()(x)
x = Activation('relu')(x)
x = GlobalMaxPooling2D()(x)
x = Dropout(dropout)(x)
x = Dense(100, activation='relu')(x)
x = Dropout(dropout)(x)
x = Dense(50, activation='relu')(x)
x = Dropout(dropout)(x)
x = Dense(10, activation='relu')(x)
x = Dropout(dropout)(x)
x = Dense(nb_classes)(x)
nvidia_net = models.Model(get_source_inputs(model_input), x)
return nvidia_net
def build_model(self, input_tensor, output_stride=16, return_model=False, alpha=1.0):
if K.image_data_format() == 'channels_last':
bn_axis = 3
else:
bn_axis = 1
if input_tensor is None:
input_tensor = Input(shape=[None, None, 3], name='input')
input_tensor_source = get_source_inputs(input_tensor)[0]
if alpha not in [0.35, 0.50, 0.75, 1.0, 1.3, 1.4]:
raise ValueError('If imagenet weights are being loaded, '
'alpha can be one of'
'`0.25`, `0.50`, `0.75` or `1.0` only.')
self.encoder = self.build_encoder(input_tensor_source, output_stride, alpha)
self.model = self.build_decoder(self.encoder, output_stride, alpha)
def VGG16(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
int_pooling='max',
classes=1000):
if weights not in {'imagenet', None}:
raise ValueError(
'The "weigths" argument should be either "None" or "imagenet"')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError(
'If using "weights" as imagenet with "include_top" as True, "classes" should be 1000'
)
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=48,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# Block 1
x = Conv2D(64, (3, 3),
activation='relu',
padding='same',
name='block1_conv1')(img_input)
x = Conv2D(64, (3, 3),
activation='relu',
padding='same',
name='block1_conv2')(x)
if int_pooling == 'avg':
x = AveragePooling2D((2, 2), strides=(2, 2), name='block1_pool')(x)
else:
x = MaxPool2D((2, 2), strides=(2, 2), name='block1_pool')(x)
# Block 2
x = Conv2D(128, (3, 3),
activation='relu',
padding='same',
name='block2_conv1')(x)
x = Conv2D(128, (3, 3),
activation='relu',
padding='same',
name='block2_conv2')(x)
if int_pooling == 'avg':
x = AveragePooling2D((2, 2), strides=(2, 2), name='block2_pool')(x)
else:
x = MaxPool2D((2, 2), strides=(2, 2), name='block2_pool')(x)
# Block 3
x = Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv1')(x)
x = Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv2')(x)
x = Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv3')(x)
if int_pooling == 'avg':
x = AveragePooling2D((2, 2), strides=(2, 2), name='block3_pool')(x)
else:
x = MaxPool2D((2, 2), strides=(2, 2), name='block3_pool')(x)
# Block 4
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv1')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv2')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv3')(x)
if int_pooling == 'avg':
x = AveragePooling2D((2, 2), strides=(2, 2), name='block4_pool')(x)
else:
x = MaxPool2D((2, 2), strides=(2, 2), name='block4_pool')(x)
# Block 5
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv1')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv2')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv3')(x)
if int_pooling == 'avg':
x = AveragePooling2D((2, 2), strides=(2, 2), name='block5_pool')(x)
else:
x = MaxPool2D((2, 2), strides=(2, 2), name='block5_pool')(x)
if include_top:
x = Flatten(name='flatten')(x)
x = Dense(4096, activation='relu', name='fc1')(x)
x = Dense(4096, activation='relu', name='fc2')(x)
x = Dense(classes, activation='softmax', name='predictions')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPool2D()(x)
# check input_shape predecessors
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
model = Model(inputs, x, name='vgg16')
# load weights
if weights == 'imagenet':
if include_top:
weights_path = get_file(
'vgg16_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
cache_dir='models')
else:
weights_path = get_file(
'vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5',
WEIGHTS_PATH_NO_TOP,
cache_dir='models')
model.load_weights(weights_path)
return model
def test_multi_input_layer():
####################################################
# test multi-input layer
a = Input(shape=(32,), name='input_a')
b = Input(shape=(32,), name='input_b')
dense = Dense(16, name='dense_1')
a_2 = dense(a)
b_2 = dense(b)
merged = layers.concatenate([a_2, b_2], name='merge')
assert merged._keras_shape == (None, 16 * 2)
merge_layer, merge_node_index, merge_tensor_index = merged._keras_history
assert merge_node_index == 0
assert merge_tensor_index == 0
assert len(merge_layer._inbound_nodes) == 1
assert len(merge_layer._outbound_nodes) == 0
assert len(merge_layer._inbound_nodes[0].input_tensors) == 2
assert len(merge_layer._inbound_nodes[0].inbound_layers) == 2
c = Dense(64, name='dense_2')(merged)
d = Dense(5, name='dense_3')(c)
model = Model(inputs=[a, b], outputs=[c, d], name='model')
assert len(model.layers) == 6
print('model.input_layers:', model.input_layers)
print('model.input_layers_node_indices:', model.input_layers_node_indices)
print('model.input_layers_tensor_indices:', model.input_layers_tensor_indices)
print('model.output_layers', model.output_layers)
print('output_shape:', model.compute_output_shape([(None, 32), (None, 32)]))
assert model.compute_output_shape([(None, 32), (None, 32)]) == [(None, 64), (None, 5)]
print('mask:', model.compute_mask([a, b], [None, None]))
assert model.compute_mask([a, b], [None, None]) == [None, None]
print('output_shape:', model.compute_output_shape([(None, 32), (None, 32)]))
assert model.compute_output_shape([(None, 32), (None, 32)]) == [(None, 64), (None, 5)]
# we don't check names of first 2 layers (inputs) because
# ordering of same-level layers is not fixed
print('layers:', [layer.name for layer in model.layers])
assert [l.name for l in model.layers][2:] == ['dense_1', 'merge', 'dense_2', 'dense_3']
print('input_layers:', [l.name for l in model.input_layers])
assert [l.name for l in model.input_layers] == ['input_a', 'input_b']
print('output_layers:', [l.name for l in model.output_layers])
assert [l.name for l in model.output_layers] == ['dense_2', 'dense_3']
# actually run model
fn = K.function(model.inputs, model.outputs)
input_a_np = np.random.random((10, 32))
input_b_np = np.random.random((10, 32))
fn_outputs = fn([input_a_np, input_b_np])
assert [x.shape for x in fn_outputs] == [(10, 64), (10, 5)]
# test get_source_inputs
print(get_source_inputs(c))
assert get_source_inputs(c) == [a, b]
# serialization / deserialization
json_config = model.to_json()
recreated_model = model_from_json(json_config)
recreated_model.compile('rmsprop', 'mse')
print('recreated:')
print([layer.name for layer in recreated_model.layers])
print([layer.name for layer in recreated_model.input_layers])
print([layer.name for layer in recreated_model.output_layers])
assert [l.name for l in recreated_model.layers][2:] == ['dense_1', 'merge', 'dense_2', 'dense_3']
assert [l.name for l in recreated_model.input_layers] == ['input_a', 'input_b']
assert [l.name for l in recreated_model.output_layers] == ['dense_2', 'dense_3']
fn = K.function(recreated_model.inputs, recreated_model.outputs)
input_a_np = np.random.random((10, 32))
input_b_np = np.random.random((10, 32))
fn_outputs = fn([input_a_np, input_b_np])
assert [x.shape for x in fn_outputs] == [(10, 64), (10, 5)]
def ResNet50(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
"""Instantiates the ResNet50 architecture.
Optionally loads weights pre-trained
on ImageNet. Note that when using TensorFlow,
for best performance you should set
`image_data_format='channels_last'` in your Keras config
at ~/.keras/keras.json.
The model and the weights are compatible with both
TensorFlow and Theano. The data format
convention used by the model is the one
specified in your Keras config file.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(224, 224, 3)` (with `channels_last` data format)
or `(3, 224, 224)` (with `channels_first` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 197.
E.g. `(200, 200, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=197,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_last':
bn_axis = 3
else:
bn_axis = 1
x = Conv2D(64, (7, 7), strides=(2, 2), padding='same',
name='conv1')(img_input)
x = BatchNormalization(axis=bn_axis, name='bn_conv1')(x)
x = Activation('relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = conv_block(x, 3, [64, 64, 256], stage=2, block='a', strides=(1, 1))
x = identity_block(x, 3, [64, 64, 256], stage=2, block='b')
x = identity_block(x, 3, [64, 64, 256], stage=2, block='c')
x = conv_block(x, 3, [128, 128, 512], stage=3, block='a')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='b')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='c')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='d')
x = conv_block(x, 3, [256, 256, 1024], stage=4, block='a')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='b')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='c')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='d')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='e')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='f')
x = conv_block(x, 3, [512, 512, 2048], stage=5, block='a')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='b')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='c')
x = AveragePooling2D((7, 7), name='avg_pool')(x)
if include_top:
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='fc1000')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='resnet50')
# load weights
if weights == 'imagenet':
if include_top:
weights_path = WEIGHTS_PATH
else:
weights_path = WEIGHTS_PATH_NO_TOP
model.load_weights(weights_path)
if K.backend() == 'theano':
layer_utils.convert_all_kernels_in_model(model)
if include_top:
maxpool = model.get_layer(name='avg_pool')
shape = maxpool.output_shape[1:]
dense = model.get_layer(name='fc1000')
layer_utils.convert_dense_weights_data_format(
dense, shape, 'channels_first')
if K.image_data_format() == 'channels_first' and K.backend(
) == 'tensorflow':
warnings.warn('You are using the TensorFlow backend, yet you '
'are using the Theano '
'image data format convention '
'(`image_data_format="channels_first"`). '
'For best performance, set '
'`image_data_format="channels_last"` in '
'your Keras config '
'at ~/.keras/keras.json.')
elif weights is not None:
model.load_weights(weights)
return model
def build_mycov(repetitions=(2, 2, 2, 2),
input_tensor=None,
input_shape=None,
classes=1,
first_filters=64):
"""
TODO
"""
if input_tensor is None:
img_input = Input(shape=input_shape, name='data')
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# get parameters for model layers
no_scale_bn_params = get_bn_params(scale=False)
conv_params = get_conv_params(padding='same')
init_filters = first_filters
# resnext bottom
x = BatchNormalization(name='bn_data', **no_scale_bn_params)(img_input)
# down
dw_h_convs = OrderedDict()
for stage, rep in enumerate(repetitions):
for block in range(rep):
filters = init_filters * (2**stage)
x = SK_change_block(filters, stage, block, behind_sknet=True)(x)
if block == 1:
dw_h_convs[stage] = x
x = MaxPooling2D((2, 2), name='pooling' + str(stage + 1))(x)
filters = init_filters * (2**4)
x = SK_change_block(filters, 4, 0, behind_sknet=True)(x)
x = SK_change_block(filters, 4, 1, behind_sknet=True)(x)
# up
for stage, rep in enumerate(repetitions):
stage += 5
filters = init_filters * (2**(4 - stage + 5 - 1))
x = Conv2DTranspose(filters, (2, 2),
strides=(2, 2),
name='up' + str(stage + 1),
padding='same')(x)
x = Concatenate(axis=-1,
name='merge' + str(stage + 1) +
str(9 - stage))([x, dw_h_convs[8 - stage]])
for block in range(rep):
x = SK_change_block(filters, stage, block, behind_sknet=False)(x)
x = Conv2D(classes, (1, 1),
activation='sigmoid',
name='output',
**conv_params)(x)
# Ensure that the model takes into account any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model
model = Model(inputs, x)
return model
def ResNet152(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
eps = 1.1e-5
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=197,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor,
shape=input_shape,
name='data')
else:
img_input = input_tensor
# Handle dimension ordering for different backends
if K.image_dim_ordering() == 'tf':
bn_axis = 3
else:
bn_axis = 1
x = ZeroPadding2D((3, 3), name='conv1_zeropadding')(img_input)
x = Conv2D(64, (7, 7), strides=(2, 2), name='conv1', use_bias=False)(x)
x = BatchNormalization(epsilon=eps, axis=bn_axis, name='bn_conv1')(x)
x = Scale(axis=bn_axis, name='scale_conv1')(x)
x = Activation('relu', name='conv1_relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2), name='pool1', padding='same')(x)
x = conv_block(x, 3, [64, 64, 256], stage=2, block='a', strides=(1, 1))
x = identity_block(x, 3, [64, 64, 256], stage=2, block='b')
x = identity_block(x, 3, [64, 64, 256], stage=2, block='c')
x = conv_block(x, 3, [128, 128, 512], stage=3, block='a')
for i in range(1, 8):
x = identity_block(x, 3, [128, 128, 512], stage=3, block='b' + str(i))
x = conv_block(x, 3, [256, 256, 1024], stage=4, block='a')
for i in range(1, 36):
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='b' + str(i))
x = conv_block(x, 3, [512, 512, 2048], stage=5, block='a')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='b')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='c')
if include_top:
# Classification block
x = AveragePooling2D((7, 7), name='avg_pool')(x)
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='fc1000')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model
model = Model(inputs, x, name='resnet152')
# Load weights
if weights == 'imagenet':
if include_top:
weights_path = get_file(
'resnet152_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
cache_subdir='models',
md5_hash='cdb18a2158b88e392c0905d47dcef965')
else:
weights_path = get_file(
'resnet152_weights_tf_dim_ordering_tf_kernels_no_top.h5',
WEIGHTS_PATH_NO_TOP,
cache_subdir='models',
md5_hash='02cb9130cc51543cd703c79697baa592')
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def ResNet50(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
"""Instantiates the ResNet50 architecturefor training the motion stream of the Temporal Segment Networks.
Optionally loads weights pre-trained on ImageNet. This model can
only be used with the data format `(width, height, channels)`.
You should set `image_data_format='channels_last'` in your Keras config
located at ~/.keras/keras.json.
Note that the default input image size for this model is 224x224.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as stacked optical flows input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_last':
bn_axis = 3
else:
bn_axis = 1
x = ZeroPadding2D(padding=(3, 3), name='conv1_pad')(img_input)
x = Conv2D(64, (7, 7), strides=(2, 2), padding='valid', name='conv1')(x)
x = BatchNormalization(axis=bn_axis, name='bn_conv1')(x)
x = Activation('relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = conv_block(x, 3, [64, 64, 256], stage=2, block='a', strides=(1, 1))
x = identity_block(x, 3, [64, 64, 256], stage=2, block='b')
x = identity_block(x, 3, [64, 64, 256], stage=2, block='c')
x = conv_block(x, 3, [128, 128, 512], stage=3, block='a')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='b')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='c')
x = identity_block(x, 3, [128, 128, 512], stage=3, block='d')
x = conv_block(x, 3, [256, 256, 1024], stage=4, block='a')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='b')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='c')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='d')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='e')
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='f')
x = conv_block(x, 3, [512, 512, 2048], stage=5, block='a')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='b')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='c')
x = AveragePooling2D((7, 7), name='avg_pool')(x)
if include_top:
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='fc1000')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='resnet50')
# load weights
if weights == 'imagenet':
if include_top:
weights_path = get_file(
'resnet50_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
cache_subdir='models',
md5_hash='a7b3fe01876f51b976af0dea6bc144eb')
else:
weights_path = get_file(
'resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5',
WEIGHTS_PATH_NO_TOP,
cache_subdir='models',
md5_hash='a268eb855778b3df3c7506639542a6af')
load_imagenet_weights(weights_path, model, input_shape)
if K.backend() == 'theano':
layer_utils.convert_all_kernels_in_model(model)
elif weights is not None:
weights_file = h5py.File(weights)
layer_names = [_ for _ in weights_file['model_weights']['resnet50']]
weights_dict = weights_file['model_weights']['resnet50']
for layer_name in layer_names:
layer = model.get_layer(layer_name)
weights = [
param.value for param in weights_dict[layer_name].values()
]
if layer_name == 'conv1':
init_kernel = np.average(weights[0], axis=2)
init_kernel = np.reshape(
init_kernel, (init_kernel.shape[0], init_kernel.shape[1],
1, init_kernel.shape[2]))
init_kernel = np.dstack([init_kernel] * input_shape[2])
if len(weights) == 1:
layer.set_weights([init_kernel])
else:
layer.set_weights([init_kernel, weights[1]])
else:
layer.set_weights(weights)
return model
def MobileNet(input_shape=None,
alpha=1.0,
depth_multiplier=1,
dropout=1e-3,
include_top=True,
weights='imagenet',
input_tensor=None,
pooling=None,
classes=1000,
channel_format='channels_last'):
"""Instantiates the MobileNet architecture.
To load a MobileNet model via `load_model`, import the custom
objects `relu6` and pass them to the `custom_objects` parameter.
E.g.
model = load_model('mobilenet.h5', custom_objects={
'relu6': mobilenet.relu6})
# Arguments
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(224, 224, 3)`
(with `channels_last` data format)
or (3, 224, 224) (with `channels_first` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 32.
E.g. `(200, 200, 3)` would be one valid value.
alpha: controls the width of the network.
- If `alpha` < 1.0, proportionally decreases the number
of filters in each layer.
- If `alpha` > 1.0, proportionally increases the number
of filters in each layer.
- If `alpha` = 1, default number of filters from the paper
are used at each layer.
depth_multiplier: depth multiplier for depthwise convolution
(also called the resolution multiplier)
dropout: dropout rate
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of
`layers.Input()`)
to use as image input for the model.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model
will be the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a
2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
RuntimeError: If attempting to run this model with a
backend that does not support separable convolutions.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as ImageNet with `include_top` '
'as true, `classes` should be 1000')
# Determine proper input shape and default size.
if input_shape is None:
default_size = 224
else:
if channel_format == 'channels_first':
rows = input_shape[1]
cols = input_shape[2]
else:
rows = input_shape[0]
cols = input_shape[1]
if rows == cols and rows in [128, 160, 192, 224]:
default_size = rows
else:
default_size = 224
input_shape = _obtain_input_shape(input_shape,
default_size=default_size,
min_size=32,
data_format=channel_format,
require_flatten=include_top,
weights=weights)
if backend.image_data_format() == 'channels_last':
row_axis, col_axis = (0, 1)
else:
row_axis, col_axis = (1, 2)
rows = input_shape[row_axis]
cols = input_shape[col_axis]
if weights == 'imagenet':
if depth_multiplier != 1:
raise ValueError('If imagenet weights are being loaded, '
'depth multiplier must be 1')
if alpha not in [0.25, 0.50, 0.75, 1.0]:
raise ValueError('If imagenet weights are being loaded, '
'alpha can be one of'
'`0.25`, `0.50`, `0.75` or `1.0` only.')
if rows != cols or rows not in [128, 160, 192, 224]:
if rows is None:
rows = 224
warnings.warn('MobileNet shape is undefined.'
' Weights for input shape '
'(224, 224) will be loaded.')
else:
raise ValueError('If imagenet weights are being loaded, '
'input must have a static square shape '
'(one of (128, 128), (160, 160), '
'(192, 192), or (224, 224)). '
'Input shape provided = %s' % (input_shape, ))
if backend.image_data_format() != 'channels_last':
warnings.warn('The MobileNet family of models is only available '
'for the input data format "channels_last" '
'(width, height, channels). '
'However your settings specify the default '
'data format "channels_first" (channels, width, height).'
' You should set `image_data_format="channels_last"` '
'in your Keras config located at ~/.keras/keras.json. '
'The model being returned right now will expect inputs '
'to follow the "channels_last" data format.')
backend.set_image_data_format('channels_last')
old_data_format = 'channels_first'
else:
old_data_format = None
if input_tensor is None:
img_input = layers.Input(shape=input_shape)
else:
if not backend.is_keras_tensor(input_tensor):
img_input = layers.Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
x = conv_block(img_input, 32, alpha, strides=(2, 2))
x = depthwise_conv_block(x, 64, alpha, depth_multiplier, block_id=1)
x = depthwise_conv_block(x,
128,
alpha,
depth_multiplier,
strides=(2, 2),
block_id=2)
x = depthwise_conv_block(x, 128, alpha, depth_multiplier, block_id=3)
x = depthwise_conv_block(x,
256,
alpha,
depth_multiplier,
strides=(2, 2),
block_id=4)
x = depthwise_conv_block(x, 256, alpha, depth_multiplier, block_id=5)
x = depthwise_conv_block(x,
512,
alpha,
depth_multiplier,
strides=(2, 2),
block_id=6)
x = depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=7)
x = depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=8)
x = depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=9)
x = depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=10)
x = depthwise_conv_block(x, 512, alpha, depth_multiplier, block_id=11)
x = depthwise_conv_block(x,
1024,
alpha,
depth_multiplier,
strides=(2, 2),
block_id=12)
x = depthwise_conv_block(x, 1024, alpha, depth_multiplier, block_id=13)
if include_top:
if backend.image_data_format() == 'channels_first':
shape = (int(1024 * alpha), 1, 1)
else:
shape = (1, 1, int(1024 * alpha))
x = layers.GlobalAveragePooling2D()(x)
x = layers.Reshape(shape, name='reshape_1')(x)
x = layers.Dropout(dropout, name='dropout')(x)
x = layers.Conv2D(classes, (1, 1), padding='same',
name='conv_preds')(x)
x = layers.Activation('softmax', name='act_softmax')(x)
x = layers.Reshape((classes, ), name='reshape_2')(x)
else:
if pooling == 'avg':
x = layers.GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = layers.GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = engine.get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = models.Model(inputs, x, name='mobilenet_%0.2f_%s' % (alpha, rows))
# load weights
if weights == 'imagenet':
if backend.image_data_format() == 'channels_first':
raise ValueError('Weights for "channels_first" format '
'are not available.')
if alpha == 1.0:
alpha_text = '1_0'
elif alpha == 0.75:
alpha_text = '7_5'
elif alpha == 0.50:
alpha_text = '5_0'
else:
alpha_text = '2_5'
if include_top:
model_name = 'mobilenet_%s_%d_tf.h5' % (alpha_text, rows)
weight_path = BASE_WEIGHT_PATH + model_name
weights_path = keras_utils.get_file(model_name,
weight_path,
cache_subdir='models')
else:
model_name = 'mobilenet_%s_%d_tf_no_top.h5' % (alpha_text, rows)
weight_path = BASE_WEIGHT_PATH + model_name
weights_path = keras_utils.get_file(model_name,
weight_path,
cache_subdir='models')
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
if old_data_format:
backend.set_image_data_format(old_data_format)
return model
def cspresnet50(include_top=False, input_shape=None, input_tensor=None, weights_file=None, classes=1000):
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=32,
data_format=K.image_data_format(),
weights=None,
require_flatten=True)
if input_tensor is None:
img_input = Input(input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_last':
bn_axis = 3
else:
bn_axis = 1
x = Conv2D(64, (7, 7), strides=(2, 2), use_bias=False, padding='same', name="conv1")(img_input)
x = BatchNormalization(axis=bn_axis, name="conv1/bn")(x)
x = LeakyReLU()(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='pool1')(x)
temp = Conv2D(128, (1, 1), strides=(1, 1), padding='same')(x)
temp = BatchNormalization(axis=bn_axis, name='conv2/bn')(temp)
temp = LeakyReLU()(temp)
x = Conv2D(64, (1, 1), strides=(1, 1), padding='same')(x)
x = BatchNormalization(name='conv3/bn')(x)
x = LeakyReLU()(x)
# Resnet stage 1
x = resnet_identity_block(x, (3, 3), [64, 64, 128], stage=1, block="a")
x = resnet_identity_block(x, (3, 3), [64, 64, 128], stage=1, block="b")
x = resnet_identity_block(x, (3, 3), [64, 64, 128], stage=1, block="c")
# Cross Stage 1
x, temp = cross_stage(x, temp, [128, 128, 128, 256, 256], stage=1)
# Resnet stage 2
x = resnet_identity_block(x, (3, 3), [128, 128, 256], stage=2, block='a')
x = resnet_identity_block(x, (3, 3), [128, 128, 256], stage=2, block='b')
x = resnet_identity_block(x, (3, 3), [128, 128, 256], stage=2, block='c')
# Cross Stage 2
x, temp = cross_stage(x, temp, [256, 256, 256, 512, 512], stage=2)
# Resnet stage 3
x = resnet_identity_block(x, (3, 3), [256, 256, 512], stage=3, block='a')
x = resnet_identity_block(x, (3, 3), [256, 256, 512], stage=3, block='b')
x = resnet_identity_block(x, (3, 3), [256, 256, 512], stage=3, block='c')
x = resnet_identity_block(x, (3, 3), [256, 256, 512], stage=3, block='d')
x = resnet_identity_block(x, (3, 3), [256, 256, 512], stage=3, block='e')
# Cross Stage 3
x, temp = cross_stage(x, temp, [512, 512, 512, 1024, 1024], stage=3)
# Resnet stage 4
x = resnet_identity_block(x, (3, 3), [512, 512, 1024], stage=4, block='a')
x = resnet_identity_block(x, (3, 3), [512, 512, 1024], stage=4, block='b')
x = Conv2D(1024, (1, 1), strides=(1, 1), padding='same', name='conv4')(x)
x = BatchNormalization(axis=bn_axis, name='conv4/bn')(x)
x = LeakyReLU()(x)
x = layers.add([x, temp])
x = Conv2D(1024, (1, 1), strides=(1, 1), padding='same', name='conv5')(x)
x = LeakyReLU()(x)
x = AveragePooling2D((7, 7), name='pool5')(x)
if include_top:
x = Flatten()(x)
x = Dense(classes, name='fc')
x = Activation('softmax')(x)
else:
x = GlobalMaxPooling2D()(x)
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
model = Model(inputs, x, name="CSPResnet")
model.summary()
if weights_file is not None:
model.load_weights(weights_file)
return model
def build_model(self,
input_tensor,
output_stride=8,
return_model=False,
alpha=1.0):
if K.image_data_format() == 'channels_last':
bn_axis = 3
else:
bn_axis = 1
if input_tensor is None:
input_tensor = Input(shape=[None, None, 3], name='input')
input_tensor_source = get_source_inputs(input_tensor)[0]
if alpha not in [0.35, 0.50, 0.75, 1.0, 1.3, 1.4]:
raise ValueError('If imagenet weights are being loaded, '
'alpha can be one of'
'`0.25`, `0.50`, `0.75` or `1.0` only.')
first_block_filters = _make_divisible(32 * alpha, 8)
x = Conv2D(first_block_filters,
kernel_size=3,
strides=(2, 2),
padding='same',
use_bias=False,
name='mn_Conv1')(input_tensor)
x = BatchNormalization(epsilon=1e-3,
momentum=0.999,
name='mn_bn_Conv1')(x)
x = Activation(tf.nn.relu6, name='mn_Conv1_relu')(x)
current_stride = 2
stride_left = output_stride / current_stride
self.conv0 = _inverted_res_block(x,
filters=16,
alpha=alpha,
stride=1,
expansion=1,
block_id=0)
if stride_left > 1:
strides = (2, 2)
dilation = 1
stride_left /= 2
else:
strides = (1, 1)
dilation = 2
self.conv1 = _inverted_res_block(self.conv0,
filters=24,
alpha=alpha,
stride=strides,
dilation=dilation,
expansion=6,
block_id=1,
use_shortcuts=True)
self.conv2 = _inverted_res_block(self.conv1,
filters=24,
alpha=alpha,
stride=1,
expansion=6,
block_id=2)
if stride_left > 1:
strides = (2, 2)
dilation = 1
stride_left /= 2
else:
strides = (1, 1)
dilation = 2
self.conv3 = _inverted_res_block(self.conv2,
filters=32,
alpha=alpha,
stride=strides,
dilation=dilation,
expansion=6,
block_id=3,
use_shortcuts=True)
self.conv4 = _inverted_res_block(self.conv3,
filters=32,
alpha=alpha,
stride=1,
expansion=6,
block_id=4,
use_shortcuts=True)
self.conv5 = _inverted_res_block(self.conv4,
filters=32,
alpha=alpha,
stride=1,
expansion=6,
block_id=5,
use_shortcuts=True)
if stride_left > 1:
strides = (2, 2)
dilation = 1
stride_left /= 2
else:
strides = (1, 1)
dilation = 2
self.conv6 = _inverted_res_block(self.conv5,
filters=64,
alpha=alpha,
stride=strides,
dilation=dilation,
expansion=6,
block_id=6,
use_shortcuts=True)
self.conv7 = _inverted_res_block(self.conv6,
filters=64,
alpha=alpha,
stride=1,
expansion=6,
block_id=7,
use_shortcuts=True)
self.conv8 = _inverted_res_block(self.conv7,
filters=64,
alpha=alpha,
stride=1,
expansion=6,
block_id=8,
use_shortcuts=True)
self.conv9 = _inverted_res_block(self.conv8,
filters=64,
alpha=alpha,
stride=1,
expansion=6,
block_id=9,
use_shortcuts=True)
self.conv10 = _inverted_res_block(self.conv9,
filters=96,
alpha=alpha,
stride=1,
expansion=6,
block_id=10,
use_shortcuts=True)
self.conv11 = _inverted_res_block(self.conv10,
filters=96,
alpha=alpha,
stride=1,
expansion=6,
block_id=11,
use_shortcuts=True)
self.conv12 = _inverted_res_block(self.conv11,
filters=96,
alpha=alpha,
stride=1,
expansion=6,
block_id=12,
use_shortcuts=True)
if stride_left > 1:
strides = (2, 2)
dilation = 1
stride_left /= 2
else:
strides = (1, 1)
dilation = 2
self.conv13 = _inverted_res_block(self.conv12,
filters=160,
alpha=alpha,
stride=strides,
dilation=dilation,
expansion=6,
block_id=13,
use_shortcuts=True)
self.conv14 = _inverted_res_block(self.conv13,
filters=160,
alpha=alpha,
stride=1,
expansion=6,
block_id=14,
use_shortcuts=True)
self.conv15 = _inverted_res_block(self.conv14,
filters=160,
alpha=alpha,
stride=1,
expansion=6,
block_id=15,
use_shortcuts=True)
self.conv16 = _inverted_res_block(self.conv15,
filters=320,
alpha=alpha,
stride=1,
expansion=6,
block_id=16,
use_shortcuts=True)
# no alpha applied to last conv as stated in the paper:
# if the width multiplier is greater than 1 we
# increase the number of output channels
if alpha > 1.0:
last_block_filters = _make_divisible(1280 * alpha, 8)
else:
last_block_filters = 1280
x = Conv2D(last_block_filters,
kernel_size=1,
use_bias=False,
name='mn_Conv_1')(self.conv16)
x = BatchNormalization(epsilon=1e-3,
momentum=0.999,
name='mn_Conv_1_bn')(x)
x = Activation(tf.nn.relu6, name='mn_out_relu')(x)
rows = 224
# Create model.
self.model = Model(input_tensor_source,
x,
name='mobilenetv2_%0.2f_%s' % (alpha, rows))
# load weights
if K.image_data_format() == 'channels_first':
raise ValueError('Weights for "channels_first" format '
'are not available.')
model_name = 'mobilenet_v2_weights_tf_dim_ordering_tf_kernels_' + \
str(alpha) + '_' + str(rows) + '_no_top' + '.h5'
weigh_path = BASE_WEIGHT_PATH + model_name
weights_path = get_file(model_name, weigh_path, cache_subdir='models')
print(f"Loading weights from {weights_path}")
self.model.load_weights(weights_path)
def VGG16(include_top=True, weights='imagenet', input_tensor=None, input_shape=None, pooling=None, classes=1000):
"""Instantiates the VGG16 architecture.
Optionally loads weights pre-trained
on ImageNet. Note that when using TensorFlow,
for best performance you should set
`image_data_format='channels_last'` in your Keras config
at ~/.keras/keras.json.
The model and the weights are compatible with both
TensorFlow and Theano. The data format
convention used by the model is the one
specified in your Keras config file.
# Arguments
include_top: whether to include the 3 fully-connected
layers at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(224, 224, 3)` (with `channels_last` data format)
or `(3, 224, 224)` (with `channels_first` data format).
It should have exactly 3 input channels,
and width and height should be no smaller than 48.
E.g. `(200, 200, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=IMG_DIM,
min_size=48,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# Block 1
x = Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv1')(img_input)
x = Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv2')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(x)
# Block 2
x = Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv1')(x)
x = Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv2')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(x)
# Block 3
x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv1')(x)
x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv2')(x)
x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool')(x)
# Block 4
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv1')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv2')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool')(x)
# Block 5
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv1')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv2')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block5_pool')(x)
if include_top:
# Classification block
x = Flatten(name='flatten')(x)
x = Dense(4096, activation='relu', name='fc1')(x)
x = Dense(4096, activation='relu', name='fc2')(x)
# x = Dense(classes, activation='softmax', name='predictions')(x) # todo
x = Dense(LABEL_DIM, activation='sigmoid', name='predictions')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='vgg16')
# load weights
if weights == 'imagenet':
if include_top:
weights_path = get_file('vgg16_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
cache_subdir='models',
file_hash='64373286793e3c8b2b4e3219cbf3544b')
else:
weights_path = get_file('vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5',
WEIGHTS_PATH_NO_TOP,
cache_subdir='models',
file_hash='6d6bbae143d832006294945121d1f1fc')
model.load_weights(weights_path)
if K.backend() == 'theano':
layer_utils.convert_all_kernels_in_model(model)
if K.image_data_format() == 'channels_first':
if include_top:
maxpool = model.get_layer(name='block5_pool')
shape = maxpool.output_shape[1:]
dense = model.get_layer(name='fc1')
layer_utils.convert_dense_weights_data_format(dense, shape, 'channels_first')
if K.backend() == 'tensorflow':
warnings.warn('You are using the TensorFlow backend, yet you '
'are using the Theano '
'image data format convention '
'(`image_data_format="channels_first"`). '
'For best performance, set '
'`image_data_format="channels_last"` in '
'your Keras config '
'at ~/.keras/keras.json.')
elif weights is not None:
model.load_weights(weights)
return model
def InceptionV3(include_top=False,
weights='imagenet',
input_tensor=None,
input_shape=(512,512,3),
pooling=None,
classes=1):
"""Instantiates the Inception v3 architecture.
Optionally loads weights pre-trained on ImageNet.
Note that the data format convention used by the model is
the one specified in your Keras config at `~/.keras/keras.json`.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(299, 299, 3)` (with `channels_last` data format)
or `(3, 299, 299)` (with `channels_first` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 139.
E.g. `(150, 150, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as `"imagenet"` with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
# input_shape = _obtain_input_shape(
# input_shape,
# default_size=299,
# min_size=139,
# data_format=backend.image_data_format(),
# require_flatten=False,
# weights=weights)
if input_tensor is None:
img_input = layers.Input(shape=input_shape)
else:
if not backend.is_keras_tensor(input_tensor):
img_input = layers.Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if backend.image_data_format() == 'channels_first':
channel_axis = 1
else:
channel_axis = 3
x = conv2d_bn(img_input, 32, 3, 3, strides=(2, 2), padding='same')
x = conv2d_bn(x, 32, 3, 3, padding='same')
x = conv2d_bn(x, 64, 3, 3)
skip1 = x
x = layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)
x = conv2d_bn(x, 80, 1, 1, padding='same')
x = conv2d_bn(x, 192, 3, 3, padding='same')
skip2 = x
x = layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)
# mixed 0, 1, 2: 35 x 35 x 256
branch1x1 = conv2d_bn(x, 64, 1, 1)
branch5x5 = conv2d_bn(x, 48, 1, 1)
branch5x5 = conv2d_bn(branch5x5, 64, 5, 5)
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch_pool = layers.AveragePooling2D((3, 3),
strides=(1, 1),
padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 32, 1, 1)
x = layers.concatenate(
[branch1x1, branch5x5, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed0')
# mixed 1: 35 x 35 x 256
branch1x1 = conv2d_bn(x, 64, 1, 1)
branch5x5 = conv2d_bn(x, 48, 1, 1)
branch5x5 = conv2d_bn(branch5x5, 64, 5, 5)
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch_pool = layers.AveragePooling2D((3, 3),
strides=(1, 1),
padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 64, 1, 1)
x = layers.concatenate(
[branch1x1, branch5x5, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed1')
# mixed 2: 35 x 35 x 256
branch1x1 = conv2d_bn(x, 64, 1, 1)
branch5x5 = conv2d_bn(x, 48, 1, 1)
branch5x5 = conv2d_bn(branch5x5, 64, 5, 5)
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch_pool = layers.AveragePooling2D((3, 3),
strides=(1, 1),
padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 64, 1, 1)
x = layers.concatenate(
[branch1x1, branch5x5, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed2')
skip3 = x
# mixed 3: 17 x 17 x 768
branch3x3 = conv2d_bn(x, 384, 3, 3, strides=(2, 2), padding='same')
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(
branch3x3dbl, 96, 3, 3, strides=(2, 2), padding='same')
# branch3x3dbl, 96, 3, 3, strides=(2, 2), padding='valid')
branch_pool = layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)
x = layers.concatenate(
[branch3x3, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed3')
# mixed 4: 17 x 17 x 768
branch1x1 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(x, 128, 1, 1)
branch7x7 = conv2d_bn(branch7x7, 128, 1, 7)
branch7x7 = conv2d_bn(branch7x7, 192, 7, 1)
branch7x7dbl = conv2d_bn(x, 128, 1, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 128, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 128, 1, 7)
branch7x7dbl = conv2d_bn(branch7x7dbl, 128, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch_pool = layers.AveragePooling2D((3, 3),
strides=(1, 1),
padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch7x7, branch7x7dbl, branch_pool],
axis=channel_axis,
name='mixed4')
# mixed 5, 6: 17 x 17 x 768
for i in range(2):
branch1x1 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(x, 160, 1, 1)
branch7x7 = conv2d_bn(branch7x7, 160, 1, 7)
branch7x7 = conv2d_bn(branch7x7, 192, 7, 1)
branch7x7dbl = conv2d_bn(x, 160, 1, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 160, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 160, 1, 7)
branch7x7dbl = conv2d_bn(branch7x7dbl, 160, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch_pool = layers.AveragePooling2D(
(3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch7x7, branch7x7dbl, branch_pool],
axis=channel_axis,
name='mixed' + str(5 + i))
# mixed 7: 17 x 17 x 768
branch1x1 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(branch7x7, 192, 1, 7)
branch7x7 = conv2d_bn(branch7x7, 192, 7, 1)
branch7x7dbl = conv2d_bn(x, 192, 1, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch_pool = layers.AveragePooling2D((3, 3),
strides=(1, 1),
padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch7x7, branch7x7dbl, branch_pool],
axis=channel_axis,
name='mixed7')
skip4 = x
# mixed 8: 8 x 8 x 1280
branch3x3 = conv2d_bn(x, 192, 1, 1)
branch3x3 = conv2d_bn(branch3x3, 320, 3, 3,
strides=(2, 2), padding='same')
# strides=(2, 2), padding='valid')
branch7x7x3 = conv2d_bn(x, 192, 1, 1)
branch7x7x3 = conv2d_bn(branch7x7x3, 192, 1, 7)
branch7x7x3 = conv2d_bn(branch7x7x3, 192, 7, 1)
branch7x7x3 = conv2d_bn(
branch7x7x3, 192, 3, 3, strides=(2, 2), padding='same')
# branch7x7x3, 192, 3, 3, strides=(2, 2), padding='valid')
branch_pool = layers.MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)
# branch_pool = layers.MaxPooling2D((3, 3), strides=(2, 2))(x)
x = layers.concatenate(
[branch3x3, branch7x7x3, branch_pool],
axis=channel_axis,
name='mixed8')
# mixed 9: 8 x 8 x 2048
for i in range(2):
branch1x1 = conv2d_bn(x, 320, 1, 1)
branch3x3 = conv2d_bn(x, 384, 1, 1)
branch3x3_1 = conv2d_bn(branch3x3, 384, 1, 3)
branch3x3_2 = conv2d_bn(branch3x3, 384, 3, 1)
branch3x3 = layers.concatenate(
[branch3x3_1, branch3x3_2],
axis=channel_axis,
name='mixed9_' + str(i))
branch3x3dbl = conv2d_bn(x, 448, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 384, 3, 3)
branch3x3dbl_1 = conv2d_bn(branch3x3dbl, 384, 1, 3)
branch3x3dbl_2 = conv2d_bn(branch3x3dbl, 384, 3, 1)
branch3x3dbl = layers.concatenate(
[branch3x3dbl_1, branch3x3dbl_2], axis=channel_axis)
branch_pool = layers.AveragePooling2D(
(3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch3x3, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed' + str(9 + i))
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = engine.get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = models.Model(inputs, x, name='inception_v3')
# Load weights.
if weights == 'imagenet':
if include_top:
weights_path = keras_utils.get_file(
'inception_v3_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
cache_subdir='models',
file_hash='9a0d58056eeedaa3f26cb7ebd46da564')
else:
weights_path = keras_utils.get_file(
'inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5',
WEIGHTS_PATH_NO_TOP,
cache_subdir='models',
file_hash='bcbd6486424b2319ff4ef7d526e38f63')
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return [x, skip4, skip3, skip2, skip1]
def mobilenetv2(include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000,
t=[1, 6, 6, 6, 6, 6, 6],
alpha=1.0):
input_shape = _obtain_input_shape(input_shape=input_shape,
default_size=224,
min_size=32,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor():
img_input = Input(shape=input_shape, tensor=input_tensor)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_last':
bn_axis = -1
else:
bn_axis = 1
x = Conv2D(32, (3, 3), strides=(2, 2), padding='same',
name='conv1')(img_input)
x = BatchNormalization(axis=bn_axis, name="conv1_bn")(x)
x = ReLU(6, name="conv1_relu")(x)
x = _residual_block(x,
16, (3, 3),
block=2,
t=t[0],
alpha=alpha,
strides=(1, 1),
n=1)
x = _residual_block(x,
24, (3, 3),
block=3,
t=t[1],
alpha=alpha,
strides=(2, 2),
n=2)
x = _residual_block(x,
32, (3, 3),
block=5,
t=t[2],
alpha=alpha,
strides=(2, 2),
n=3)
x = _residual_block(x,
64, (3, 3),
block=8,
t=t[3],
alpha=alpha,
strides=(2, 2),
n=4)
x = _residual_block(x,
96, (3, 3),
block=12,
t=t[4],
alpha=alpha,
strides=(1, 1),
n=3)
x = _residual_block(x,
160, (3, 3),
block=15,
t=t[5],
alpha=alpha,
strides=(2, 2),
n=3)
x = _residual_block(x,
1280, (3, 3),
block=18,
t=t[6],
alpha=alpha,
strides=(1, 1),
n=1)
x = Conv2D(1280, (1, 1), strides=(1, 1), name="conv2")(x)
x = BatchNormalization(axis=bn_axis, name="conv2_bn")(x)
x = ReLU(6, name="conv2_relu")(x)
x = AveragePooling2D((7, 7), name="avgpool")(x)
if include_top:
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='classifier')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
if input_tensor is None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
model = Model(inputs, x, name='moblienetv2')
model.summary()
return model
def mobilenet(include_top=True,
weights="imagenet",
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
input_shape = _obtain_input_shape(input_shape=input_shape,
default_size=224,
min_size=32,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor():
img_input = Input(shape=input_shape, tensor=input_tensor)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_last':
bn_axis = -1
else:
bn_axis = 1
# Input(224, 224, 3)
# Output(112, 112, 64)
x = Conv2D(32, (3, 3), strides=(2, 2), padding='same',
name='conv1')(img_input)
x = BatchNormalization(axis=bn_axis, name="conv1_bn")(x)
x = ReLU(6, name="conv1_relu")(x)
x = _depthwise_separable_Conv(x,
64,
block=2,
bn_axis=bn_axis,
strides=(1, 1))
# Input(112, 112, 64)
# Output(56, 56, 128)
x = _depthwise_separable_Conv(x,
128,
block=3,
bn_axis=bn_axis,
strides=(2, 2))
x = _depthwise_separable_Conv(x,
128,
block=4,
bn_axis=bn_axis,
strides=(1, 1))
# Input(56, 56, 128)
# Output(28, 28, 256)
x = _depthwise_separable_Conv(x,
256,
block=5,
bn_axis=bn_axis,
strides=(2, 2))
x = _depthwise_separable_Conv(x,
256,
block=6,
bn_axis=bn_axis,
strides=(1, 1))
x = DepthwiseConv2D((3, 3), strides=(1, 1), padding='same',
name='conv3_3')(x)
# Input(28, 28, 256)
# Output(14, 14, 512)
x = _depthwise_separable_Conv(x,
512,
block=7,
bn_axis=bn_axis,
strides=(2, 2))
for i in range(8, 13, 2):
x = _depthwise_separable_Conv(x,
512,
block=i,
bn_axis=bn_axis,
strides=(1, 1))
# Input(14, 14, 612)
# Output(7, 7, 1024)
x = _depthwise_separable_Conv(x,
1024,
block=13,
bn_axis=bn_axis,
strides=(2, 2))
x = _depthwise_separable_Conv(x,
1024,
block=14,
bn_axis=bn_axis,
strides=(1, 1))
x = AveragePooling2D((7, 7), name='pool')(x)
if include_top:
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='classifier')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
if input_tensor is None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
model = Model(inputs, x, name='moblienet')
model.summary()
return model
def Xception(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
"""Instantiates the Xception architecture.
Optionally loads weights pre-trained
on ImageNet. This model is available for TensorFlow only,
and can only be used with inputs following the TensorFlow
data format `(width, height, channels)`.
You should set `image_data_format='channels_last'` in your Keras config
located at ~/.keras/keras.json.
Note that the default input image size for this model is 299x299.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(299, 299, 3)`.
It should have exactly 3 inputs channels,
and width and height should be no smaller than 71.
E.g. `(150, 150, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
RuntimeError: If attempting to run this model with a
backend that does not support separable convolutions.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
if K.backend() != 'tensorflow':
raise RuntimeError('The Xception model is only available with '
'the TensorFlow backend.')
if K.image_data_format() != 'channels_last':
warnings.warn(
'The Xception model is only available for the '
'input data format "channels_last" '
'(width, height, channels). '
'However your settings specify the default '
'data format "channels_first" (channels, width, height). '
'You should set `image_data_format="channels_last"` in your Keras '
'config located at ~/.keras/keras.json. '
'The model being returned right now will expect inputs '
'to follow the "channels_last" data format.')
K.set_image_data_format('channels_last')
old_data_format = 'channels_first'
else:
old_data_format = None
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=299,
min_size=71,
data_format=K.image_data_format(),
require_flatten=False,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
x = Conv2D(32, (3, 3), strides=(2, 2), use_bias=False,
name='block1_conv1')(img_input)
x = BatchNormalization(name='block1_conv1_bn')(x)
x = Activation('relu', name='block1_conv1_act')(x)
x = Conv2D(64, (3, 3), use_bias=False, name='block1_conv2')(x)
x = BatchNormalization(name='block1_conv2_bn')(x)
x = Activation('relu', name='block1_conv2_act')(x)
residual = Conv2D(128, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = SeparableConv2D(128, (3, 3),
padding='same',
use_bias=False,
name='block2_sepconv1')(x)
x = BatchNormalization(name='block2_sepconv1_bn')(x)
x = Activation('relu', name='block2_sepconv2_act')(x)
x = SeparableConv2D(128, (3, 3),
padding='same',
use_bias=False,
name='block2_sepconv2')(x)
x = BatchNormalization(name='block2_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block2_pool')(x)
x = layers.add([x, residual])
residual = Conv2D(256, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = Activation('relu', name='block3_sepconv1_act')(x)
x = SeparableConv2D(256, (3, 3),
padding='same',
use_bias=False,
name='block3_sepconv1')(x)
x = BatchNormalization(name='block3_sepconv1_bn')(x)
x = Activation('relu', name='block3_sepconv2_act')(x)
x = SeparableConv2D(256, (3, 3),
padding='same',
use_bias=False,
name='block3_sepconv2')(x)
x = BatchNormalization(name='block3_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block3_pool')(x)
x = layers.add([x, residual])
residual = Conv2D(728, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = Activation('relu', name='block4_sepconv1_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name='block4_sepconv1')(x)
x = BatchNormalization(name='block4_sepconv1_bn')(x)
x = Activation('relu', name='block4_sepconv2_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name='block4_sepconv2')(x)
x = BatchNormalization(name='block4_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block4_pool')(x)
x = layers.add([x, residual])
for i in range(8):
residual = x
prefix = 'block' + str(i + 5)
x = Activation('relu', name=prefix + '_sepconv1_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name=prefix + '_sepconv1')(x)
x = BatchNormalization(name=prefix + '_sepconv1_bn')(x)
x = Activation('relu', name=prefix + '_sepconv2_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name=prefix + '_sepconv2')(x)
x = BatchNormalization(name=prefix + '_sepconv2_bn')(x)
x = Activation('relu', name=prefix + '_sepconv3_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name=prefix + '_sepconv3')(x)
x = BatchNormalization(name=prefix + '_sepconv3_bn')(x)
x = layers.add([x, residual])
residual = Conv2D(1024, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = Activation('relu', name='block13_sepconv1_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name='block13_sepconv1')(x)
x = BatchNormalization(name='block13_sepconv1_bn')(x)
x = Activation('relu', name='block13_sepconv2_act')(x)
x = SeparableConv2D(1024, (3, 3),
padding='same',
use_bias=False,
name='block13_sepconv2')(x)
x = BatchNormalization(name='block13_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block13_pool')(x)
x = layers.add([x, residual])
x = SeparableConv2D(1536, (3, 3),
padding='same',
use_bias=False,
name='block14_sepconv1')(x)
x = BatchNormalization(name='block14_sepconv1_bn')(x)
x = Activation('relu', name='block14_sepconv1_act')(x)
x = SeparableConv2D(2048, (3, 3),
padding='same',
use_bias=False,
name='block14_sepconv2')(x)
x = BatchNormalization(name='block14_sepconv2_bn')(x)
x = Activation('relu', name='block14_sepconv2_act')(x)
if include_top:
x = GlobalAveragePooling2D(name='avg_pool')(x)
x = Dense(classes, activation='softmax', name='predictions')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='xception')
# load weights
if weights == 'imagenet':
if include_top:
weights_path = WEIGHTS_PATH
else:
weights_path = WEIGHTS_PATH_NO_TOP
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
if old_data_format:
K.set_image_data_format(old_data_format)
return model
def build_resnet(repetitions=(2, 2, 2, 2),
include_top=True,
input_tensor=None,
input_shape=None,
classes=1000,
block_type='conv',
attention=None):
"""
TODO
"""
if input_tensor is None:
img_input = Input(shape=input_shape, name='data')
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# choose residual block type
if block_type == 'conv':
residual_block = residual_conv_block
elif block_type == 'bottleneck':
residual_block = residual_bottleneck_block
else:
raise ValueError(
'Block type "{}" not in ["conv", "bottleneck"]'.format(block_type))
# choose attention block type
if attention == 'sse':
attention_block = SpatialSE()
elif attention == 'cse':
attention_block = ChannelSE(reduction=16)
elif attention == 'csse':
attention_block = ChannelSpatialSE(reduction=2)
elif attention is None:
attention_block = None
else:
raise ValueError(
'Supported attention blocks are: sse, cse, csse. Got "{}".'.format(
attention))
# get parameters for model layers
no_scale_bn_params = get_bn_params(scale=False)
bn_params = get_bn_params()
conv_params = get_conv_params()
init_filters = 64
# resnet bottom
x = BatchNormalization(name='bn_data', **no_scale_bn_params)(img_input)
x = ZeroPadding2D(padding=(3, 3))(x)
x = Conv2D(init_filters, (7, 7),
strides=(2, 2),
name='conv0',
**conv_params)(x)
x = BatchNormalization(name='bn0', **bn_params)(x)
x = Activation('relu', name='relu0')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = MaxPooling2D((3, 3), strides=(2, 2), padding='valid',
name='pooling0')(x)
# resnet body
for stage, rep in enumerate(repetitions):
for block in range(rep):
filters = init_filters * (2**stage)
# first block of first stage without strides because we have maxpooling before
if block == 0 and stage == 0:
x = residual_block(filters,
stage,
block,
strides=(1, 1),
cut='post',
attention=attention_block)(x)
elif block == 0:
x = residual_block(filters,
stage,
block,
strides=(2, 2),
cut='post',
attention=attention_block)(x)
else:
x = residual_block(filters,
stage,
block,
strides=(1, 1),
cut='pre',
attention=attention_block)(x)
x = BatchNormalization(name='bn1', **bn_params)(x)
x = Activation('relu', name='relu1')(x)
# resnet top
if include_top:
x = GlobalAveragePooling2D(name='pool1')(x)
x = Dense(classes, name='fc1')(x)
x = Activation('softmax', name='softmax')(x)
# Ensure that the model takes into account any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x)
return model
def Xception(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
"""Instantiates the Xception architecture for training the motion stream of the Temporal Segment Networks.
Optionally loads weights pre-trained on ImageNet or UCF101. This model can
only be used with the data format `(width, height, channels)`.
You should set `image_data_format='channels_last'` in your Keras config
located at ~/.keras/keras.json.
Note that the default input image size for this model is 299x299.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
RuntimeError: If attempting to run this model with a
backend that does not support separable convolutions.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
if K.image_data_format() != 'channels_last':
warnings.warn(
'The Xception model is only available for the '
'input data format "channels_last" '
'(width, height, channels). '
'However your settings specify the default '
'data format "channels_first" (channels, width, height). '
'You should set `image_data_format="channels_last"` in your Keras '
'config located at ~/.keras/keras.json. '
'The model being returned right now will expect inputs '
'to follow the "channels_last" data format.')
K.set_image_data_format('channels_last')
old_data_format = 'channels_first'
else:
old_data_format = None
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
x = Conv2D(32, (3, 3), strides=(2, 2), use_bias=False,
name='block1_conv1')(img_input)
x = BatchNormalization(name='block1_conv1_bn')(x)
x = Activation('relu', name='block1_conv1_act')(x)
x = Conv2D(64, (3, 3), use_bias=False, name='block1_conv2')(x)
x = BatchNormalization(name='block1_conv2_bn')(x)
x = Activation('relu', name='block1_conv2_act')(x)
residual = Conv2D(128, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = SeparableConv2D(128, (3, 3),
padding='same',
use_bias=False,
name='block2_sepconv1')(x)
x = BatchNormalization(name='block2_sepconv1_bn')(x)
x = Activation('relu', name='block2_sepconv2_act')(x)
x = SeparableConv2D(128, (3, 3),
padding='same',
use_bias=False,
name='block2_sepconv2')(x)
x = BatchNormalization(name='block2_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block2_pool')(x)
x = layers.add([x, residual])
residual = Conv2D(256, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = Activation('relu', name='block3_sepconv1_act')(x)
x = SeparableConv2D(256, (3, 3),
padding='same',
use_bias=False,
name='block3_sepconv1')(x)
x = BatchNormalization(name='block3_sepconv1_bn')(x)
x = Activation('relu', name='block3_sepconv2_act')(x)
x = SeparableConv2D(256, (3, 3),
padding='same',
use_bias=False,
name='block3_sepconv2')(x)
x = BatchNormalization(name='block3_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block3_pool')(x)
x = layers.add([x, residual])
residual = Conv2D(728, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = Activation('relu', name='block4_sepconv1_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name='block4_sepconv1')(x)
x = BatchNormalization(name='block4_sepconv1_bn')(x)
x = Activation('relu', name='block4_sepconv2_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name='block4_sepconv2')(x)
x = BatchNormalization(name='block4_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block4_pool')(x)
x = layers.add([x, residual])
for i in range(8):
residual = x
prefix = 'block' + str(i + 5)
x = Activation('relu', name=prefix + '_sepconv1_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name=prefix + '_sepconv1')(x)
x = BatchNormalization(name=prefix + '_sepconv1_bn')(x)
x = Activation('relu', name=prefix + '_sepconv2_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name=prefix + '_sepconv2')(x)
x = BatchNormalization(name=prefix + '_sepconv2_bn')(x)
x = Activation('relu', name=prefix + '_sepconv3_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name=prefix + '_sepconv3')(x)
x = BatchNormalization(name=prefix + '_sepconv3_bn')(x)
x = layers.add([x, residual])
residual = Conv2D(1024, (1, 1),
strides=(2, 2),
padding='same',
use_bias=False)(x)
residual = BatchNormalization()(residual)
x = Activation('relu', name='block13_sepconv1_act')(x)
x = SeparableConv2D(728, (3, 3),
padding='same',
use_bias=False,
name='block13_sepconv1')(x)
x = BatchNormalization(name='block13_sepconv1_bn')(x)
x = Activation('relu', name='block13_sepconv2_act')(x)
x = SeparableConv2D(1024, (3, 3),
padding='same',
use_bias=False,
name='block13_sepconv2')(x)
x = BatchNormalization(name='block13_sepconv2_bn')(x)
x = MaxPooling2D((3, 3),
strides=(2, 2),
padding='same',
name='block13_pool')(x)
x = layers.add([x, residual])
x = SeparableConv2D(1536, (3, 3),
padding='same',
use_bias=False,
name='block14_sepconv1')(x)
x = BatchNormalization(name='block14_sepconv1_bn')(x)
x = Activation('relu', name='block14_sepconv1_act')(x)
x = SeparableConv2D(2048, (3, 3),
padding='same',
use_bias=False,
name='block14_sepconv2')(x)
x = BatchNormalization(name='block14_sepconv2_bn')(x)
x = Activation('relu', name='block14_sepconv2_act')(x)
if include_top:
x = GlobalAveragePooling2D(name='avg_pool')(x)
x = Dense(classes, activation='softmax', name='predictions')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='xception')
# use cross-modality scheme to initialize weights of the first conv. layer,
# load pre-trained weights to the remaining layers as usual
if weights == 'imagenet':
if include_top:
weights_path = get_file(
'xception_weights_tf_dim_ordering_tf_kernels.h5',
TF_WEIGHTS_PATH,
cache_subdir='models',
file_hash='0a58e3b7378bc2990ea3b43d5981f1f6')
else:
weights_path = get_file(
'xception_weights_tf_dim_ordering_tf_kernels_notop.h5',
TF_WEIGHTS_PATH_NO_TOP,
cache_subdir='models',
file_hash='b0042744bf5b25fce3cb969f33bebb97')
load_imagenet_weights(weights_path, model, input_shape)
if K.backend() == 'theano':
layer_utils.convert_all_kernels_in_model(model)
elif weights is not None:
weights_file = h5py.File(weights)
layer_names = [_ for _ in weights_file['model_weights']['xception']]
weights_dict = weights_file['model_weights']['xception']
for layer_name in layer_names:
layer = model.get_layer(layer_name)
weights = [
param.value for param in weights_dict[layer_name].values()
]
if layer_name == 'block1_conv1':
init_kernel = np.average(weights[0], axis=2)
init_kernel = np.reshape(
init_kernel, (init_kernel.shape[0], init_kernel.shape[1],
1, init_kernel.shape[2]))
init_kernel = np.dstack([init_kernel] * input_shape[2])
if len(weights) == 1:
layer.set_weights([init_kernel])
else:
layer.set_weights([init_kernel, weights[1]])
else:
layer.set_weights(weights)
if old_data_format:
K.set_image_data_format(old_data_format)
return model
def ResNet152(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
""" Instantiates the ResNet152 architecture.
Optionally loads weights pre-trained
on ImageNet. Note that when using TensorFlow,
for best performance you should set
`image_data_format='channels_last'` in your Keras config
at ~/.keras/keras.json.
The model and the weights are compatible only with
TensorFlow. The data format
convention used by the model is the one
specified in your Keras config file.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(224, 224, 3)` (with `channels_last` data format)
or `(3, 224, 224)` (with `channels_first` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 197.
E.g. `(200, 200, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
eps = 1.1e-5
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=197,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor,
shape=input_shape,
name='data')
else:
img_input = input_tensor
# Handle dimension ordering for different backends
if K.image_dim_ordering() == 'tf':
bn_axis = 3
else:
bn_axis = 1
x = ZeroPadding2D((3, 3), name='conv1_zeropadding')(img_input)
x = Conv2D(64, (7, 7), strides=(2, 2), name='conv1', use_bias=False)(x)
x = BatchNormalization(epsilon=eps, axis=bn_axis, name='bn_conv1')(x)
x = Scale(axis=bn_axis, name='scale_conv1')(x)
x = Activation('relu', name='conv1_relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2), name='pool1', padding='same')(x)
x = conv_block(x, 3, [64, 64, 256], stage=2, block='a', strides=(1, 1))
x = identity_block(x, 3, [64, 64, 256], stage=2, block='b')
x = identity_block(x, 3, [64, 64, 256], stage=2, block='c')
x = conv_block(x, 3, [128, 128, 512], stage=3, block='a')
for i in range(1, 8):
x = identity_block(x, 3, [128, 128, 512], stage=3, block='b' + str(i))
x = conv_block(x, 3, [256, 256, 1024], stage=4, block='a')
for i in range(1, 36):
x = identity_block(x, 3, [256, 256, 1024], stage=4, block='b' + str(i))
x = conv_block(x, 3, [512, 512, 2048], stage=5, block='a')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='b')
x = identity_block(x, 3, [512, 512, 2048], stage=5, block='c')
if include_top:
# Classification block
x = AveragePooling2D((7, 7), name='avg_pool')(x)
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='fc1000')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model
model = Model(inputs, x, name='resnet152')
# Load weights
if weights == 'imagenet':
if include_top:
weights_path = get_file(
os.environ["MODEL_PATH"] +
'resnet152_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
md5_hash='cdb18a2158b88e392c0905d47dcef965')
else:
weights_path = get_file(
os.environ["MODEL_PATH"] +
'resnet152_weights_tf_dim_ordering_tf_kernels_no_top.h5',
WEIGHTS_PATH_NO_TOP,
md5_hash='02cb9130cc51543cd703c79697baa592')
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def InceptionResNetV2(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
"""Instantiates the Inception-ResNet v2 architecture.
Optionally loads weights pre-trained on ImageNet.
Note that the data format convention used by the model is
the one specified in your Keras config at `~/.keras/keras.json`.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is `False` (otherwise the input shape
has to be `(299, 299, 3)` (with `'channels_last'` data format)
or `(3, 299, 299)` (with `'channels_first'` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 139.
E.g. `(150, 150, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the last convolutional layer.
- `'avg'` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `'max'` means that global max pooling will be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is `True`, and
if no `weights` argument is specified.
# Returns
A Keras `Model` instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as `"imagenet"` with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
input_shape = _obtain_input_shape(
input_shape,
default_size=299,
min_size=139,
data_format=backend.image_data_format(),
require_flatten=False,
weights=weights)
if input_tensor is None:
img_input = layers.Input(shape=input_shape)
else:
if not backend.is_keras_tensor(input_tensor):
img_input = layers.Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# Stem block: 35 x 35 x 192
x = conv2d_bn(img_input, 32, 3, strides=2, padding='valid')
x = conv2d_bn(x, 32, 3, padding='valid')
x = conv2d_bn(x, 64, 3)
x = layers.MaxPooling2D(3, strides=2)(x)
x = conv2d_bn(x, 80, 1, padding='valid')
x = conv2d_bn(x, 192, 3, padding='valid')
x = layers.MaxPooling2D(3, strides=2)(x)
# Mixed 5b (Inception-A block): 35 x 35 x 320
branch_0 = conv2d_bn(x, 96, 1)
branch_1 = conv2d_bn(x, 48, 1)
branch_1 = conv2d_bn(branch_1, 64, 5)
branch_2 = conv2d_bn(x, 64, 1)
branch_2 = conv2d_bn(branch_2, 96, 3)
branch_2 = conv2d_bn(branch_2, 96, 3)
branch_pool = layers.AveragePooling2D(3, strides=1, padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 64, 1)
branches = [branch_0, branch_1, branch_2, branch_pool]
channel_axis = 1 if backend.image_data_format() == 'channels_first' else 3
x = layers.Concatenate(axis=channel_axis, name='mixed_5b')(branches)
# 10x block35 (Inception-ResNet-A block): 35 x 35 x 320
for block_idx in range(1, 11):
x = inception_resnet_block(x,
scale=0.17,
block_type='block35',
block_idx=block_idx)
# Mixed 6a (Reduction-A block): 17 x 17 x 1088
branch_0 = conv2d_bn(x, 384, 3, strides=2, padding='valid')
branch_1 = conv2d_bn(x, 256, 1)
branch_1 = conv2d_bn(branch_1, 256, 3)
branch_1 = conv2d_bn(branch_1, 384, 3, strides=2, padding='valid')
branch_pool = layers.MaxPooling2D(3, strides=2, padding='valid')(x)
branches = [branch_0, branch_1, branch_pool]
x = layers.Concatenate(axis=channel_axis, name='mixed_6a')(branches)
# 20x block17 (Inception-ResNet-B block): 17 x 17 x 1088
for block_idx in range(1, 21):
x = inception_resnet_block(x,
scale=0.1,
block_type='block17',
block_idx=block_idx)
# Mixed 7a (Reduction-B block): 8 x 8 x 2080
branch_0 = conv2d_bn(x, 256, 1)
branch_0 = conv2d_bn(branch_0, 384, 3, strides=2, padding='valid')
branch_1 = conv2d_bn(x, 256, 1)
branch_1 = conv2d_bn(branch_1, 288, 3, strides=2, padding='valid')
branch_2 = conv2d_bn(x, 256, 1)
branch_2 = conv2d_bn(branch_2, 288, 3)
branch_2 = conv2d_bn(branch_2, 320, 3, strides=2, padding='valid')
branch_pool = layers.MaxPooling2D(3, strides=2, padding='valid')(x)
branches = [branch_0, branch_1, branch_2, branch_pool]
x = layers.Concatenate(axis=channel_axis, name='mixed_7a')(branches)
# 10x block8 (Inception-ResNet-C block): 8 x 8 x 2080
for block_idx in range(1, 10):
x = inception_resnet_block(x,
scale=0.2,
block_type='block8',
block_idx=block_idx)
x = inception_resnet_block(x,
scale=1.,
activation=None,
block_type='block8',
block_idx=10)
# Final convolution block: 8 x 8 x 1536
x = conv2d_bn(x, 1536, 1, name='conv_7b')
if include_top:
# Classification block
x = layers.GlobalAveragePooling2D(name='avg_pool')(x)
x = layers.Dense(classes, name='logits')(x)
x = Activation('softmax', name='predictions')(x)
else:
if pooling == 'avg':
x = layers.GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = layers.GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = engine.get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = models.Model(inputs, x, name='inception_resnet_v2')
# Load weights.
if weights == 'imagenet':
if include_top:
fname = 'inception_resnet_v2_weights_tf_dim_ordering_tf_kernels.h5'
weights_path = keras_utils.get_file(
fname,
BASE_WEIGHT_URL + fname,
cache_subdir='models',
file_hash='e693bd0210a403b3192acc6073ad2e96')
else:
fname = ('inception_resnet_v2_weights_'
'tf_dim_ordering_tf_kernels_notop.h5')
weights_path = keras_utils.get_file(
fname,
BASE_WEIGHT_URL + fname,
cache_subdir='models',
file_hash='d19885ff4a710c122648d3b5c3b684e4')
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def NASNetA(include_top=True,
input_tensor=None,
input_shape=None,
pooling=None,
add_aux_output=False,
stem=None,
stem_filters=96,
num_cell_repeats=18,
penultimate_filters=768,
num_classes=10,
num_reduction_cells=2,
dropout_rate=0.5):
aux_outputs = None
if input_tensor is None:
input_tensor = Input(input_shape)
if pooling is None:
pooling = 'avg'
if stem is None:
stem = ImagenetStem
filters = int(penultimate_filters / ((2 ** num_reduction_cells) * 6))
prev, cur = stem(filters=filters, stem_filters=stem_filters)(input_tensor)
for repeat in range(num_reduction_cells + 1):
if repeat == num_reduction_cells and add_aux_output:
prefix = 'aux_{}'.format(repeat * num_cell_repeats - 1)
aux_outputs = AuxiliaryTop(num_classes, prefix=prefix)(cur)
if repeat > 0:
filters *= 2
prev, cur = cur, prev
cur = ReductionCell(filters, prefix='reduction_cell_{}'.format(repeat - 1))(cur, prev)
for cell_index in range(num_cell_repeats):
prev, cur = cur, prev
cur = NormalCell(filters, prefix='cell_{}'.format(cell_index + repeat * num_cell_repeats))(cur, prev)
with K.name_scope('final_layer'):
x = Activation('relu', name='last_relu')(cur)
if include_top:
x = GlobalAveragePooling2D(name='avg_pool')(x)
x = Dropout(rate=dropout_rate)(x)
outputs = Dense(num_classes, activation='softmax', name='final_layer/FC')(x)
model_suffix = 'with_top'
else:
if pooling == 'avg':
outputs = GlobalAveragePooling2D(name='avg_pool')(x)
elif pooling == 'max':
outputs = GlobalMaxPooling2D(name='max_pool')(x)
else:
outputs = None
raise Exception('Supported options for pooling: `avg` or `max` given pooling: {}'.format(pooling))
model_suffix = 'no_top'
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = input_tensor
model_name = 'NASNet-A_{}@{}_{}_{}'.format(num_cell_repeats, penultimate_filters, num_classes, model_suffix)
if add_aux_output:
return Model(inputs, [outputs, aux_outputs], name='{}_with_auxiliary_output'.format(model_name))
else:
return Model(inputs, outputs, name=model_name)
def ResNet50(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000,
layers=50):
"""Instantiates the ResNet50 architecture.
Optionally loads weights pre-trained on ImageNet.
Note that the data format convention used by the model is
the one specified in your Keras config at `~/.keras/keras.json`.
When using TensorFlow, for best performance you should
set `"image_data_format": "channels_last"` in the config.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(224, 224, 3)` (with `channels_last` data format)
or `(3, 224, 224)` (with `channels_first` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 197.
E.g. `(200, 200, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
assert layers in [18, 34, 50, 101, 152]
use_bn = (layers == 50)
basic = (layers in [18, 34])
if layers == 18:
num_layers = [2, 2, 2, 2]
elif layers == 34:
num_layers = [3, 4, 6, 3]
elif layers == 50:
num_layers = [3, 4, 6, 3]
elif layers == 101:
num_layers = [3, 4, 23, 3]
elif layers == 152:
num_layers = [3, 8, 36, 3]
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=197,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_last':
bn_axis = 3
else:
bn_axis = 1
if basic:
x = Conv2D(64, (7, 7), strides=(2, 2), padding='same',
name='conv1')(img_input)
x = Activation('relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2), padding='same')(x)
x = ResNetBlock(3, [64, 64, 256],
stage=2,
block='a',
use_bn=use_bn,
basic=basic)(x)
else:
x = ZeroPadding2D(padding=(3, 3), name='conv1_pad')(img_input)
x = Conv2D(64, (7, 7), strides=(2, 2), padding='valid',
name='conv1')(x)
if use_bn:
x = BatchNormalization(axis=bn_axis, name='bn_conv1')(x)
x = Activation('relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = ResNetBlock(3, [64, 64, 256],
stage=2,
block='a',
strides=(1, 1),
use_bn=use_bn,
basic=basic)(x)
for i in range(num_layers[0] - 1):
x = ResNetBlock(3, [64, 64, 256],
stage=2,
block=chr(ord('b') + i),
identity=True,
use_bn=use_bn,
basic=basic)(x)
x = ResNetBlock(3, [128, 128, 512],
stage=3,
block='a',
use_bn=use_bn,
basic=basic)(x)
for i in range(num_layers[1] - 1):
x = ResNetBlock(3, [128, 128, 512],
stage=3,
block=chr(ord('b') + i),
identity=True,
use_bn=use_bn,
basic=basic)(x)
x = ResNetBlock(3, [256, 256, 1024],
stage=4,
block='a',
use_bn=use_bn,
basic=basic)(x)
for i in range(num_layers[2] - 1):
x = ResNetBlock(3, [256, 256, 1024],
stage=4,
block=chr(ord('b') + i),
identity=True,
use_bn=use_bn,
basic=basic)(x)
x = ResNetBlock(3, [512, 512, 2048],
stage=5,
block='a',
use_bn=use_bn,
basic=basic)(x)
for i in range(num_layers[3] - 1):
x = ResNetBlock(3, [512, 512, 2048],
stage=5,
block=chr(ord('b') + i),
identity=True,
use_bn=use_bn,
basic=basic)(x)
if basic:
x = GlobalAveragePooling2D()(x)
else:
x = AveragePooling2D((7, 7), name='avg_pool')(x)
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='fc1000')(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='resnet50')
# load weights
if weights == 'imagenet' and layers == 50:
if include_top:
weights_path = get_file(
'resnet50_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
cache_subdir='models',
md5_hash='a7b3fe01876f51b976af0dea6bc144eb')
else:
weights_path = get_file(
'resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5',
WEIGHTS_PATH_NO_TOP,
cache_subdir='models',
md5_hash='a268eb855778b3df3c7506639542a6af')
with h5py.File(weights_path, mode='r') as f:
if 'layer_names' not in f.attrs and 'model_weights' in f:
f = f['model_weights']
import itertools
all_layers = [
[l] if not isinstance(l, ResNetBlock) else l.get_layers()
for l in model.layers
]
all_layers = list(itertools.chain.from_iterable(all_layers))
load_weights_from_hdf5_group_by_name(f, all_layers)
if K.backend() == 'theano':
layer_utils.convert_all_kernels_in_model(model)
return model
def InceptionV3(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
"""Instantiates the Inception v3 architecture.
Optionally loads weights pre-trained
on ImageNet. Note that when using TensorFlow,
for best performance you should set
`image_data_format='channels_last'` in your Keras config
at ~/.keras/keras.json.
The model and the weights are compatible with both
TensorFlow and Theano. The data format
convention used by the model is the one
specified in your Keras config file.
Note that the default input image size for this model is 299x299.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(299, 299, 3)` (with `channels_last` data format)
or `(3, 299, 299)` (with `channels_first` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 139.
E.g. `(150, 150, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
input_shape = _obtain_input_shape(
input_shape,
default_size=299,
min_size=139,
data_format=K.image_data_format(),
require_flatten=False,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_first':
channel_axis = 1
else:
channel_axis = 3
x = conv2d_bn(img_input, 32, 3, 3, strides=(2, 2), padding='valid')
x = conv2d_bn(x, 32, 3, 3, padding='valid')
x = conv2d_bn(x, 64, 3, 3)
x = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = conv2d_bn(x, 80, 1, 1, padding='valid')
x = conv2d_bn(x, 192, 3, 3, padding='valid')
x = MaxPooling2D((3, 3), strides=(2, 2))(x)
# mixed 0, 1, 2: 35 x 35 x 256
branch1x1 = conv2d_bn(x, 64, 1, 1)
branch5x5 = conv2d_bn(x, 48, 1, 1)
branch5x5 = conv2d_bn(branch5x5, 64, 5, 5)
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch_pool = AveragePooling2D((3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 32, 1, 1)
x = layers.concatenate(
[branch1x1, branch5x5, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed0')
# mixed 1: 35 x 35 x 256
branch1x1 = conv2d_bn(x, 64, 1, 1)
branch5x5 = conv2d_bn(x, 48, 1, 1)
branch5x5 = conv2d_bn(branch5x5, 64, 5, 5)
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch_pool = AveragePooling2D((3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 64, 1, 1)
x = layers.concatenate(
[branch1x1, branch5x5, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed1')
# mixed 2: 35 x 35 x 256
branch1x1 = conv2d_bn(x, 64, 1, 1)
branch5x5 = conv2d_bn(x, 48, 1, 1)
branch5x5 = conv2d_bn(branch5x5, 64, 5, 5)
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch_pool = AveragePooling2D((3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 64, 1, 1)
x = layers.concatenate(
[branch1x1, branch5x5, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed2')
# mixed 3: 17 x 17 x 768
branch3x3 = conv2d_bn(x, 384, 3, 3, strides=(2, 2), padding='valid')
branch3x3dbl = conv2d_bn(x, 64, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 96, 3, 3)
branch3x3dbl = conv2d_bn(
branch3x3dbl, 96, 3, 3, strides=(2, 2), padding='valid')
branch_pool = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = layers.concatenate(
[branch3x3, branch3x3dbl, branch_pool], axis=channel_axis, name='mixed3')
# mixed 4: 17 x 17 x 768
branch1x1 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(x, 128, 1, 1)
branch7x7 = conv2d_bn(branch7x7, 128, 1, 7)
branch7x7 = conv2d_bn(branch7x7, 192, 7, 1)
branch7x7dbl = conv2d_bn(x, 128, 1, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 128, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 128, 1, 7)
branch7x7dbl = conv2d_bn(branch7x7dbl, 128, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch_pool = AveragePooling2D((3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch7x7, branch7x7dbl, branch_pool],
axis=channel_axis,
name='mixed4')
# mixed 5, 6: 17 x 17 x 768
for i in range(2):
branch1x1 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(x, 160, 1, 1)
branch7x7 = conv2d_bn(branch7x7, 160, 1, 7)
branch7x7 = conv2d_bn(branch7x7, 192, 7, 1)
branch7x7dbl = conv2d_bn(x, 160, 1, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 160, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 160, 1, 7)
branch7x7dbl = conv2d_bn(branch7x7dbl, 160, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch_pool = AveragePooling2D(
(3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch7x7, branch7x7dbl, branch_pool],
axis=channel_axis,
name='mixed' + str(5 + i))
# mixed 7: 17 x 17 x 768
branch1x1 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(x, 192, 1, 1)
branch7x7 = conv2d_bn(branch7x7, 192, 1, 7)
branch7x7 = conv2d_bn(branch7x7, 192, 7, 1)
branch7x7dbl = conv2d_bn(x, 192, 1, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 7, 1)
branch7x7dbl = conv2d_bn(branch7x7dbl, 192, 1, 7)
branch_pool = AveragePooling2D((3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch7x7, branch7x7dbl, branch_pool],
axis=channel_axis,
name='mixed7')
# mixed 8: 8 x 8 x 1280
branch3x3 = conv2d_bn(x, 192, 1, 1)
branch3x3 = conv2d_bn(branch3x3, 320, 3, 3,
strides=(2, 2), padding='valid')
branch7x7x3 = conv2d_bn(x, 192, 1, 1)
branch7x7x3 = conv2d_bn(branch7x7x3, 192, 1, 7)
branch7x7x3 = conv2d_bn(branch7x7x3, 192, 7, 1)
branch7x7x3 = conv2d_bn(
branch7x7x3, 192, 3, 3, strides=(2, 2), padding='valid')
branch_pool = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = layers.concatenate(
[branch3x3, branch7x7x3, branch_pool], axis=channel_axis, name='mixed8')
# mixed 9: 8 x 8 x 2048
for i in range(2):
branch1x1 = conv2d_bn(x, 320, 1, 1)
branch3x3 = conv2d_bn(x, 384, 1, 1)
branch3x3_1 = conv2d_bn(branch3x3, 384, 1, 3)
branch3x3_2 = conv2d_bn(branch3x3, 384, 3, 1)
branch3x3 = layers.concatenate(
[branch3x3_1, branch3x3_2], axis=channel_axis, name='mixed9_' + str(i))
branch3x3dbl = conv2d_bn(x, 448, 1, 1)
branch3x3dbl = conv2d_bn(branch3x3dbl, 384, 3, 3)
branch3x3dbl_1 = conv2d_bn(branch3x3dbl, 384, 1, 3)
branch3x3dbl_2 = conv2d_bn(branch3x3dbl, 384, 3, 1)
branch3x3dbl = layers.concatenate(
[branch3x3dbl_1, branch3x3dbl_2], axis=channel_axis)
branch_pool = AveragePooling2D(
(3, 3), strides=(1, 1), padding='same')(x)
branch_pool = conv2d_bn(branch_pool, 192, 1, 1)
x = layers.concatenate(
[branch1x1, branch3x3, branch3x3dbl, branch_pool],
axis=channel_axis,
name='mixed' + str(9 + i))
if include_top:
# Classification block
x = GlobalAveragePooling2D(name='avg_pool')(x)
x = Dense(classes, activation='softmax', name='predictions')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='inception_v3')
# load weights
if weights == 'imagenet':
if K.image_data_format() == 'channels_first':
if K.backend() == 'tensorflow':
warnings.warn('You are using the TensorFlow backend, yet you '
'are using the Theano '
'image data format convention '
'(`image_data_format="channels_first"`). '
'For best performance, set '
'`image_data_format="channels_last"` in '
'your Keras config '
'at ~/.keras/keras.json.')
if include_top:
weights_path = WEIGHTS_PATH
else:
weights_path = WEIGHTS_PATH_NO_TOP
model.load_weights(weights_path)
elif weights is not None:
model.load_weights(weights)
return model
def build_vgg16(include_top=True,
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=48,
data_format=K.image_data_format(),
include_top=include_top)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# Block 1
x = Conv2D(64, (3, 3),
activation='relu',
padding='same',
name='block1_conv1')(img_input)
x = Conv2D(64, (3, 3),
activation='relu',
padding='same',
name='block1_conv2')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(x)
# Block 2
x = Conv2D(128, (3, 3),
activation='relu',
padding='same',
name='block2_conv1')(x)
x = Conv2D(128, (3, 3),
activation='relu',
padding='same',
name='block2_conv2')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(x)
# Block 3
x = Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv1')(x)
x = Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv2')(x)
x = Conv2D(256, (3, 3),
activation='relu',
padding='same',
name='block3_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool')(x)
# Block 4
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv1')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv2')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block4_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool')(x)
# Block 5
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv1')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv2')(x)
x = Conv2D(512, (3, 3),
activation='relu',
padding='same',
name='block5_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block5_pool')(x)
if include_top:
# Classification block
x = Flatten(name='flatten')(x)
x = Dense(4096, activation='relu', name='fc1')(x)
x = Dense(4096, activation='relu', name='fc2')(x)
x = Dense(classes, activation='softmax', name='predictions')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='vgg16')
# load weights
if weights == 'imagenet':
if include_top:
weights_path = get_file(
'vgg16_weights_tf_dim_ordering_tf_kernels.h5',
WEIGHTS_PATH,
cache_subdir='models')
else:
weights_path = get_file(
'vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5',
WEIGHTS_PATH_NO_TOP,
cache_subdir='models')
model.load_weights(weights_path)
if K.backend() == 'theano':
layer_utils.convert_all_kernels_in_model(model)
if K.image_data_format() == 'channels_first':
if include_top:
maxpool = model.get_layer(name='block5_pool')
shape = maxpool.output_shape[1:]
dense = model.get_layer(name='fc1')
layer_utils.convert_dense_weights_data_format(
dense, shape, 'channels_first')
if K.backend() == 'tensorflow':
warnings.warn('You are using the TensorFlow backend, yet you '
'are using the Theano '
'image data format convention '
'(`image_data_format="channels_first"`). '
'For best performance, set '
'`image_data_format="channels_last"` in '
'your Keras config '
'at ~/.keras/keras.json.')
return model
def build_resnext(repetitions=(2, 2, 2, 2),
include_top=True,
input_tensor=None,
input_shape=None,
classes=1000,
first_conv_filters=64,
first_block_filters=64):
"""
TODO
"""
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=97,
data_format='channels_last',
require_flatten=include_top)
if input_tensor is None:
img_input = Input(shape=input_shape, name='data')
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# get parameters for model layers
no_scale_bn_params = get_bn_params(scale=False)
bn_params = get_bn_params()
conv_params = get_conv_params()
init_filters = first_block_filters
# resnext bottom
x = BatchNormalization(name='bn_data', **no_scale_bn_params)(img_input)
x = ZeroPadding2D(padding=(3, 3))(x)
x = Conv2D(first_conv_filters, (7, 7), name='conv0', **conv_params)(x)
x = BatchNormalization(name='bn0', **bn_params)(x)
x = Activation('relu', name='relu0')(x)
x = ZeroPadding2D(padding=(1, 1))(x)
x = MaxPooling2D((3, 3), strides=(2, 2), padding='valid',
name='pooling0')(x)
# resnext body
for stage, rep in enumerate(repetitions):
for block in range(rep):
filters = init_filters * (2**stage)
# first block of first stage without strides because we have maxpooling before
if stage == 0 and block == 0:
x = conv_block(filters, stage, block, strides=(1, 1))(x)
elif block == 0:
x = conv_block(filters, stage, block, strides=(2, 2))(x)
else:
x = identity_block(filters, stage, block)(x)
# resnext top
if include_top:
x = GlobalAveragePooling2D(name='pool1')(x)
x = Dense(classes, name='fc1')(x)
x = Activation('softmax', name='softmax')(x)
# Ensure that the model takes into account any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model
model = Model(inputs, x)
return model
def build_resnet(block_count=5,
include_top=True,
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000):
# Determine proper input shape
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=197,
data_format='channels_last',
require_flatten=include_top)
if input_tensor is None:
img_input = Input(shape=input_shape, name='data')
else:
if not is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# Stage 0
x = ZeroPadding2D((3, 3))(img_input)
x = Conv2D(64, (7, 7), strides=(2, 2), name='conv1', use_bias=True)(x)
x = BatchNormalization(name='bn_conv1')(x)
x = Activation('relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2), padding="same")(x)
# Stage 1
x = conv_block(x, 3, [64, 64, 256], stage=1, block='a', strides=(1, 1))
x = identity_block(x, 3, [64, 64, 256], stage=1, block='b')
x = identity_block(x, 3, [64, 64, 256], stage=1, block='c')
# Stage 2
x = conv_block(x, 3, [128, 128, 512], stage=2, block='a')
x = identity_block(x, 3, [128, 128, 512], stage=2, block='b')
x = identity_block(x, 3, [128, 128, 512], stage=2, block='c')
x = identity_block(x, 3, [128, 128, 512], stage=2, block='d')
# Stage 3
x = conv_block(x, 3, [256, 256, 1024], stage=3, block='a')
for i in range(block_count):
x = identity_block(x, 3, [256, 256, 1024], stage=3, block=chr(98 + i))
# Stage 4
x = conv_block(x, 3, [512, 512, 2048], stage=4, block='a')
x = identity_block(x, 3, [512, 512, 2048], stage=4, block='b')
x = identity_block(x, 3, [512, 512, 2048], stage=4, block='c')
# resnet top
if include_top:
x = AveragePooling2D((7, 7), name='avg_pool')(x)
x = Dropout(0.3)(x)
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='fc1000')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# Ensure that the model takes into account any potential predecessors of `input_tensor`.
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x)
return model
def ResNet18(include_top=True, weights=None,
input_tensor=None, input_shape=(192, 160),
pooling=None,
classes=2):
"""Instantiates the ResNet50 architecture.
Optionally loads weights pre-trained
on ImageNet. Note that when using TensorFlow,
for best performance you should set
`image_data_format='channels_last'` in your Keras config
at ~/.keras/keras.json.
The model and the weights are compatible with both
TensorFlow and Theano. The data format
convention used by the model is the one
specified in your Keras config file.
# Arguments
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization),
'imagenet' (pre-training on ImageNet),
or the path to the weights file to be loaded.
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(224, 224, 3)` (with `channels_last` data format)
or `(3, 224, 224)` (with `channels_first` data format).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 197.
E.g. `(200, 200, 3)` would be one valid value.
pooling: Optional pooling mode for feature extraction
when `include_top` is `False`.
- `None` means that the output of the model will be
the 4D tensor output of the
last convolutional layer.
- `avg` means that global average pooling
will be applied to the output of the
last convolutional layer, and thus
the output of the model will be a 2D tensor.
- `max` means that global max pooling will
be applied.
classes: optional number of classes to classify images
into, only to be specified if `include_top` is True, and
if no `weights` argument is specified.
# Returns
A Keras model instance.
# Raises
ValueError: in case of invalid argument for `weights`,
or invalid input shape.
"""
if not (weights in {'imagenet', None} or os.path.exists(weights)):
raise ValueError('The `weights` argument should be either '
'`None` (random initialization), `imagenet` '
'(pre-training on ImageNet), '
'or the path to the weights file to be loaded.')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
# Determine proper input shape
input_shape = (input_shape[0], input_shape[1], 1)
# input_shape = _obtain_input_shape(input_shape, default_size=224, min_size=160, data_format=K.image_data_format(),
# require_flatten=include_top, weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if K.image_data_format() == 'channels_last':
bn_axis = -1
else:
bn_axis = 1
print(type(img_input))
x = ZeroPadding2D(padding=(3, 3), name='conv1_pad')(img_input)
x = Conv2D(64, (7, 7), strides=(2, 2), kernel_initializer='he_normal', padding='valid', name='conv1')(x)
x = BatchNormalization(axis=bn_axis, name='bn_conv1')(x)
x = LeakyReLU()(x)
x = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = convolut_block(x, 3, [64, 64], stage=2, block='a', strides=(1, 1))
x = identity_block(x, 3, [64, 64], stage=2, block='b')
x = convolut_block(x, 3, [128, 128], stage=3, block='a')
x = identity_block(x, 3, [128, 128], stage=3, block='b')
x = convolut_block(x, 3, [256, 256], stage=4, block='a')
x = identity_block(x, 3, [256, 256], stage=4, block='b')
x = convolut_block(x, 3, [512, 512], stage=5, block='a')
x = identity_block(x, 3, [512, 512], stage=5, block='b')
x = AveragePooling2D((6, 5), name='avg_pool')(x)
x = Flatten()(x)
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# Create model.
model = Model(inputs, x, name='resnet50')
return model
def resnext50(include_top=True,
weights='vggface',
input_tensor=None,
input_shape=None,
pooling=None,
classes=8631,
weight_decay=5e-4):
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=32,
data_format=K.image_data_format(),
require_flatten=include_top,
weights=weights)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
if K.image_data_format() == "channels_last":
bn_axis = 3
else:
bn_axis = 1
x = Conv2D(64, (7, 7),
use_bias=False,
strides=(2, 2),
padding='same',
kernel_initializer='he_normal',
kernel_regularizer=l2(weight_decay),
name='conv1/7x7_s2')(img_input)
x = BatchNormalization(axis=bn_axis, name='conv1/7x7_s2/bn')(x)
x = Activation('relu')(x)
x = MaxPooling2D((3, 3), strides=(2, 2))(x)
x = resnet_block(x, (3, 3), [128, 128, 256],
stage=2,
block=1,
strides=(1, 1))
x = identity_block(x, (3, 3), [128, 128, 256], stage=2, block=2)
x = identity_block(x, (3, 3), [128, 128, 256], stage=2, block=3)
x = resnet_block(x, (3, 3), [256, 256, 512], stage=3, block=1)
x = identity_block(x, (3, 3), [256, 256, 512], stage=3, block=2)
x = identity_block(x, (3, 3), [256, 256, 512], stage=3, block=3)
x = identity_block(x, (3, 3), [256, 256, 512], stage=3, block=4)
x = resnet_block(x, (3, 3), [512, 512, 1024], stage=4, block=1)
x = identity_block(x, (3, 3), [512, 512, 1024], stage=4, block=2)
x = identity_block(x, (3, 3), [512, 512, 1024], stage=4, block=3)
x = identity_block(x, (3, 3), [512, 512, 1024], stage=4, block=4)
x = identity_block(x, (3, 3), [512, 512, 1024], stage=4, block=5)
x = identity_block(x, (3, 3), [512, 512, 1024], stage=4, block=6)
x = resnet_block(x, (3, 3), [1024, 1024, 2048], stage=5, block=1)
x = identity_block(x, (3, 3), [1024, 1024, 2048], stage=5, block=2)
x = identity_block(x, (3, 3), [1024, 1024, 2048], stage=5, block=3)
x = AveragePooling2D((7, 7), name='pool5')(x)
if include_top:
x = Flatten()(x)
x = Dense(classes, activation='softmax', name='classifier')(x)
else:
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
model = Model(inputs, x, name='resnext50')
model.summary()
return model
