def call(self, u_vecs, **kwargs):
if self.share_weights:
u_hat_vecs = K.conv1d(u_vecs, self.W)
else:
u_hat_vecs = K.local_conv1d(u_vecs, self.W, [1], [1])
batch_size = K.shape(u_vecs)[0]
input_num_capsule = K.shape(u_vecs)[1]
u_hat_vecs = K.reshape(u_hat_vecs,
(batch_size, input_num_capsule,
self.num_capsule, self.dim_capsule))
u_hat_vecs = K.permute_dimensions(u_hat_vecs, (0, 2, 1, 3))
# final u_hat_vecs.shape = [None, num_capsule, input_num_capsule, dim_capsule]
b = K.zeros_like(
u_hat_vecs[:, :, :,
0]) # shape = [None, num_capsule, input_num_capsule]
for i in range(self.routings):
c = softmax(b, 1)
o = K.batch_dot(c, u_hat_vecs, [2, 2])
if K.backend() == 'theano':
o = K.sum(o, axis=1)
if i < self.routings - 1:
o = K.l2_normalize(o, -1)
b = K.batch_dot(o, u_hat_vecs, [2, 3])
if K.backend() == 'theano':
b = K.sum(b, axis=1)
return self.activation(o)
def call(self, inputs):
if K.backend() == "tensorflow":
# only do range padding in width dimension
out = K.concatenate([
inputs[:, :, self.padding:, :], inputs[:, :, :, :],
inputs[:, :, :self.padding - 1, :]
],
axis=2)
# print(K.int_shape(out)) # 我想看一下输出的形状 (None, 1, 719, 128)
else:
raise Exception("Backend " + K.backend() + "not implemented")
return out
def wrapper(*args, **kwargs):
for data_format in {'channels_first', 'channels_last'}:
K.set_image_data_format(data_format)
func(*args, **kwargs)
if K.backend() == 'tensorflow':
K.clear_session()
tf.reset_default_graph()
def save_weights_to_hdf5_group(f, layers):
"""Saves the weights of a list of layers to a HDF5 group.
Arguments:
f: HDF5 group.
layers: List of layer instances.
"""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
save_attributes_to_hdf5_group(
f, 'layer_names', [layer.name.encode('utf8') for layer in layers])
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['keras_version'] = str(keras_version).encode('utf8')
for layer in layers:
g = f.create_group(layer.name)
weights = _legacy_weights(layer)
weight_values = K.batch_get_value(weights)
weight_names = [w.name.encode('utf8') for w in weights]
save_attributes_to_hdf5_group(g, 'weight_names', weight_names)
for name, val in zip(weight_names, weight_values):
param_dset = g.create_dataset(name, val.shape, dtype=val.dtype)
if not val.shape:
# scalar
param_dset[()] = val
else:
param_dset[:] = val
def call(self, inputs):
# dilated pooling for tensorflow backend
if K.backend() == 'theano':
Exception(
'This version of DeepCell only works with the tensorflow backend'
)
df = 'NCDHW' if self.data_format == 'channels_first' else 'NDHWC'
padding_input = self.padding.upper()
if not isinstance(self.dilation_rate, tuple):
dilation_rate = (self.dilation_rate, self.dilation_rate,
self.dilation_rate)
else:
dilation_rate = self.dilation_rate
output = tf.nn.pool(inputs,
window_shape=self.pool_size,
pooling_type='MAX',
padding=padding_input,
dilation_rate=dilation_rate,
strides=self.strides,
data_format=df)
return output
def training_phase():
mean_batch = K.mean(mean_instance, axis=0, keepdims=True)
variance_batch = K.mean(temp, axis=0,
keepdims=True) - K.square(mean_batch)
mean_batch_reshaped = K.flatten(mean_batch)
variance_batch_reshaped = K.flatten(variance_batch)
if K.backend() != 'cntk':
sample_size = K.prod(
[K.shape(inputs)[axis] for axis in reduction_axes])
sample_size = K.cast(sample_size, dtype=K.dtype(inputs))
# sample variance - unbiased estimator of population variance
variance_batch_reshaped *= sample_size / (sample_size -
(1.0 + self.epsilon))
self.add_update([
K.moving_average_update(self.moving_mean, mean_batch_reshaped,
self.momentum),
K.moving_average_update(self.moving_variance,
variance_batch_reshaped, self.momentum)
], inputs)
return normalize_func(mean_batch, variance_batch)
def main():
"""Save famous example models in Keras format."""
if len(sys.argv) != 2:
print('usage: [output dir]')
sys.exit(1)
else:
assert K.backend() == "tensorflow"
assert K.floatx() == "float32"
assert K.image_data_format() == 'channels_last'
dir_path = sys.argv[1]
save_xception(dir_path, 'xception')
save_vgg16(dir_path, 'vgg16')
save_vgg19(dir_path, 'vgg19')
save_resnet50(dir_path, 'resnet50')
save_inceptionv3(dir_path, 'inceptionv3')
save_densenet201(dir_path, 'densenet201')
save_nasnetlarge(dir_path, 'nasnetlarge')
keras.applications.densenet.DenseNet201(include_top=True,
weights='imagenet',
input_tensor=None,
input_shape=None,
pooling=None,
classes=1000)
keras.applications.nasnet.NASNetLarge(input_shape=None,
include_top=True,
weights='imagenet',
input_tensor=None,
pooling=None,
classes=1000)
def _pooling_function(self, inputs, pool_size, dilation_rate, strides,
padding, data_format):
backend = K.backend()
#dilated pooling for tensorflow backend
if backend == "theano":
Exception(
'This version of DeepCell only works with the tensorflow backend'
)
if data_format == 'channels_first':
df = 'NCHW'
elif data_format == 'channel_last':
df = 'NHWC'
if self.padding == "valid":
padding_input = "VALID"
if self.padding == "same":
padding_input = "SAME"
output = tf.nn.pool(inputs,
window_shape=pool_size,
pooling_type="MAX",
padding=padding_input,
dilation_rate=(dilation_rate, dilation_rate),
strides=strides,
data_format=df)
return output
def save_weights_to_hdf5_group(f, layers):
"""Saves the weights of a list of layers to a HDF5 group.
Arguments:
f: HDF5 group.
layers: List of layer instances.
"""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
save_attributes_to_hdf5_group(
f, 'layer_names', [layer.name.encode('utf8') for layer in layers])
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['keras_version'] = str(keras_version).encode('utf8')
for layer in layers:
g = f.create_group(layer.name)
weights = _legacy_weights(layer)
weight_values = K.batch_get_value(weights)
weight_names = [w.name.encode('utf8') for w in weights]
save_attributes_to_hdf5_group(g, 'weight_names', weight_names)
for name, val in zip(weight_names, weight_values):
param_dset = g.create_dataset(name, val.shape, dtype=val.dtype)
if not val.shape:
# scalar
param_dset[()] = val
else:
param_dset[:] = val
def call(self, inputs):
backend = K.backend()
if backend == "theano":
Exception(
'This version of DeepCell only works with the tensorflow backend'
)
if self.data_format == 'channels_first':
output = tf.tensordot(inputs, self.kernel, axes=[[1], [0]])
output = tf.transpose(output, perm=[0, 3, 1, 2])
# output = K.dot(inputs, self.kernel)
elif self.data_format == 'channels_last':
output = tf.tensordot(inputs, self.kernel, axes=[[3], [0]])
if self.use_bias:
output = K.bias_add(output,
self.bias,
data_format=self.data_format)
if self.activation is not None:
return self.activation(output)
return output
def load_weights(model):
get_file = tf.contrib.keras.utils.get_file
if K.backend() == "theano":
if K.image_data_format() == "channels_first":
weights = get_file(
'inception_v4_weights_th_dim_ordering_th_kernels.h5',
TH_BACKEND_TH_DIM_ORDERING,
cache_subdir='models')
else:
weights = get_file(
'inception_v4_weights_tf_dim_ordering_th_kernels.h5',
TH_BACKEND_TF_DIM_ORDERING,
cache_subdir='models')
else:
if K.image_data_format() == "channels_first":
weights = get_file(
'inception_v4_weights_th_dim_ordering_tf_kernels.h5',
TF_BACKEND_TH_DIM_ORDERING,
cache_subdir='models')
else:
weights = get_file(
'inception_v4_weights_tf_dim_ordering_tf_kernels.h5',
TH_BACKEND_TF_DIM_ORDERING,
cache_subdir='models')
model.load_weights(weights, by_name=True)
print("Model weights loaded.", weights)
def save_weights_to_hdf5_group(f, layers):
"""Saves the weights of a list of layers to a HDF5 group.
Arguments:
f: HDF5 group.
layers: List of layer instances.
"""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
save_attributes_to_hdf5_group(
f, 'layer_names', [layer.name.encode('utf8') for layer in layers])
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['keras_version'] = str(keras_version).encode('utf8')
for layer in layers:
g = f.create_group(layer.name)
symbolic_weights = layer.weights
weight_values = K.batch_get_value(symbolic_weights)
weight_names = []
for i, (w, val) in enumerate(zip(symbolic_weights, weight_values)):
if hasattr(w, 'name') and w.name:
name = str(w.name)
else:
name = 'param_' + str(i)
weight_names.append(name.encode('utf8'))
save_attributes_to_hdf5_group(g, 'weight_names', weight_names)
for name, val in zip(weight_names, weight_values):
param_dset = g.create_dataset(name, val.shape, dtype=val.dtype)
if not val.shape:
# scalar
param_dset[()] = val
else:
param_dset[:] = val
def _load_state_dict(self, model, weight_dict):
original_keras_version = keras_version
original_backend = K.backend()
weight_value_tuples = []
for k, layer in enumerate(model.layers):
weight_names = [l.name for l in layer.weights]
if len(weight_names) == 0:
continue
weight_values = [
np.asarray(weight_dict[weight_name])
for weight_name in weight_names
]
symbolic_weights = layer.trainable_weights + layer.non_trainable_weights
weight_values = preprocess_weights_for_loading(
layer, weight_values, original_keras_version, original_backend)
if len(weight_values) != len(symbolic_weights):
raise ValueError("Layer #" + str(k) + ' (named "' +
layer.name +
'" in the current model) was found to '
"correspond to layer " + layer.name +
" in the save file. "
"However the new layer " + layer.name +
" expects " + str(len(symbolic_weights)) +
" weights, but the saved weights have " +
str(len(weight_values)) + " elements.")
weight_value_tuples += zip(symbolic_weights, weight_values)
K.batch_set_value(weight_value_tuples)
def main():
"""Generate different test models and save them to the given directory."""
if len(sys.argv) != 3:
print('usage: [model name] [destination file path]')
sys.exit(1)
else:
model_name = sys.argv[1]
dest_path = sys.argv[2]
get_model_functions = {
'small': get_test_model_small,
'sequential': get_test_model_sequential,
'full': get_test_model_full
}
if not model_name in get_model_functions:
print('unknown model name: ', model_name)
sys.exit(2)
assert K.backend() == "tensorflow"
assert K.floatx() == "float32"
assert K.image_data_format() == 'channels_last'
np.random.seed(0)
model_func = get_model_functions[model_name]
model = model_func()
model.save(dest_path, include_optimizer=False)
# Make sure models can be loaded again,
# see https://github.com/fchollet/keras/issues/7682
model = load_model(dest_path)
print(model.summary())
def save_weights_to_hdf5_group(f, layers):
"""Saves the weights of a list of layers to a HDF5 group.
Arguments:
f: HDF5 group.
layers: List of layer instances.
"""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
save_attributes_to_hdf5_group(
f, 'layer_names', [layer.name.encode('utf8') for layer in layers])
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['keras_version'] = str(keras_version).encode('utf8')
for layer in layers:
g = f.create_group(layer.name)
symbolic_weights = layer.weights
weight_values = K.batch_get_value(symbolic_weights)
weight_names = []
for i, (w, val) in enumerate(zip(symbolic_weights, weight_values)):
if hasattr(w, 'name') and w.name:
name = str(w.name)
else:
name = 'param_' + str(i)
weight_names.append(name.encode('utf8'))
save_attributes_to_hdf5_group(g, 'weight_names', weight_names)
for name, val in zip(weight_names, weight_values):
param_dset = g.create_dataset(name, val.shape, dtype=val.dtype)
if not val.shape:
# scalar
param_dset[()] = val
else:
param_dset[:] = val
def batchnorm(x, momentum=0.9, epsilon=1e-5, name=None):
"""
Batch normalization layer.
Parameters:
----------
x : keras.backend tensor/variable/symbol
Input tensor/variable/symbol.
momentum : float, default 0.9
Momentum for the moving average.
epsilon : float, default 1e-5
Small float added to variance to avoid dividing by zero.
name : str, default None
Layer name.
Returns
-------
keras.backend tensor/variable/symbol
Resulted tensor/variable/symbol.
"""
if K.backend() == "mxnet":
x = GluonBatchNormalization(momentum=momentum,
epsilon=epsilon,
name=name)(x)
else:
x = nn.BatchNormalization(axis=get_channel_axis(),
momentum=momentum,
epsilon=epsilon,
name=name)(x)
return x
def call(self, inputs, mask=None):
pos = K.relu(inputs)
if K.backend() == 'theano':
neg = (K.pattern_broadcast(self.alpha, self.param_broadcast) *
(inputs - math_ops.abs(inputs)) * 0.5)
else:
neg = -self.alpha * K.relu(-inputs)
return pos + neg
def model_metadata(model, include_optimizer=True, require_config=True):
"""Returns a dictionary containing the model metadata."""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
from tensorflow.python.keras.optimizer_v2 import optimizer_v2 # pylint: disable=g-import-not-at-top
model_config = {'class_name': model.__class__.__name__}
try:
model_config['config'] = model.get_config()
except NotImplementedError as e:
if require_config:
raise e
metadata = dict(keras_version=str(keras_version),
backend=K.backend(),
model_config=model_config)
if model.optimizer and include_optimizer:
if isinstance(model.optimizer, optimizers.TFOptimizer):
logging.warning(
'TensorFlow optimizers do not '
'make it possible to access '
'optimizer attributes or optimizer state '
'after instantiation. '
'As a result, we cannot save the optimizer '
'as part of the model save file. '
'You will have to compile your model again after loading it. '
'Prefer using a Keras optimizer instead '
'(see keras.io/optimizers).')
else:
try:
metadata['training_config'] = {
'loss': model.loss,
# pylint: disable=protected-access
'metrics': model._compile_metrics,
'weighted_metrics': model._compile_weighted_metrics,
# pylint: enable=protected-access
'sample_weight_mode': model.sample_weight_mode,
'loss_weights': model.loss_weights,
}
if isinstance(model.optimizer, optimizer_v2.RestoredOptimizer):
raise NotImplementedError(
'As of now, Optimizers loaded from SavedModel cannot be saved. '
'If you\'re calling `model.save` or `tf.keras.models.save_model`,'
' please set the `include_optimizer` option to `False`. For '
'`tf.saved_model.save`, delete the optimizer from the model.'
)
else:
optimizer_config = {
'class_name': model.optimizer.__class__.__name__,
'config': model.optimizer.get_config()
}
metadata['training_config'][
'optimizer_config'] = optimizer_config
except AttributeError:
pass # If the model has an optimizer, but not all of the attributes
# loss, _compile_metrics, etc., then it was not compiled using
# model.compile. In this case, do not save the training config.
return metadata
def call(self, inputs, mask=None):
pos = K.relu(inputs)
if K.backend() == 'theano':
neg = (
K.pattern_broadcast(self.alpha, self.param_broadcast) *
(inputs - math_ops.abs(inputs)) * 0.5)
else:
neg = -self.alpha * K.relu(-inputs)
return pos + neg
def __init__(self, scale=2, data_format=None, **kwargs):
super(Resize, self).__init__(**kwargs)
backend = K.backend()
if backend == "theano":
Exception(
'This version of DeepCell only works with the tensorflow backend'
)
self.data_format = conv_utils.normalize_data_format(data_format)
self.scale = scale
def call(self, inputs, **kwargs):
padding = self.padding
pool_size = self.pool_size
strides = self.strides
if K.backend() == 'tensorflow':
ksize = [1, pool_size[0], pool_size[1], 1]
padding = padding.upper()
strides = [1, strides[0], strides[1], 1]
output, argmax = tf.nn.max_pool_with_argmax(inputs,
ksize=ksize,
strides=strides,
padding=padding)
else:
errmsg = '{} backend is not supported for layer {}'.format(
K.backend(),
type(self).__name__)
raise NotImplementedError(errmsg)
argmax = K.cast(argmax, K.floatx())
return [output, argmax]
def convert(in_path, out_path):
"""Convert any Keras model to the frugally-deep model format."""
assert K.backend() == "tensorflow"
assert K.floatx() == "float32"
assert K.image_data_format() == 'channels_last'
print('loading {}'.format(in_path))
with CustomObjectScope({
'relu6': mobilenet.relu6,
'DepthwiseConv2D': mobilenet.DepthwiseConv2D
}):
model = load_model(in_path)
# Force creation of underlying functional model.
# see: https://github.com/fchollet/keras/issues/8136
# Loss and optimizer type do not matter, since to don't train the model.
model.compile(loss='mse', optimizer='sgd')
model = convert_sequential_to_model(model)
test_data = gen_test_data(model)
json_output = {}
json_output['architecture'] = json.loads(model.to_json())
json_output['image_data_format'] = K.image_data_format()
for depth in range(1, 3, 1):
json_output['conv2d_valid_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_conv2d_eval, 'valid')
json_output['conv2d_same_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_conv2d_eval, 'same')
json_output['separable_conv2d_valid_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_sep_conv2d_eval, 'valid')
json_output['separable_conv2d_same_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_sep_conv2d_eval, 'same')
json_output['max_pooling_2d_valid_offset'] =\
check_operation_offset(1, conv2d_offset_max_pool_eval, 'valid')
json_output['max_pooling_2d_same_offset'] =\
check_operation_offset(1, conv2d_offset_max_pool_eval, 'same')
json_output['average_pooling_2d_valid_offset'] =\
check_operation_offset(1, conv2d_offset_average_pool_eval, 'valid')
json_output['average_pooling_2d_same_offset'] =\
check_operation_offset(1, conv2d_offset_average_pool_eval, 'same')
json_output['input_shapes'] = get_shapes(test_data['inputs'])
json_output['output_shapes'] = get_shapes(test_data['outputs'])
json_output['tests'] = [test_data]
json_output['trainable_params'] = get_all_weights(model)
print('writing {}'.format(out_path))
write_text_file(
out_path,
json.dumps(json_output, allow_nan=False, indent=2, sort_keys=True))
def dot_product(x, kernel):
"""
Wrapper for dot product operation, in order to be compatible with both
Theano and Tensorflow
Args:
x (): input
kernel (): weights
Returns:
"""
if K.backend() == 'tensorflow':
return K.squeeze(K.dot(x, K.expand_dims(kernel)), axis=-1)
else:
return K.dot(x, kernel)
def _time_distributed_dense(x,
w,
b=None,
dropout=None,
input_dim=None,
output_dim=None,
timesteps=None,
training=None):
"""Apply `y . w + b` for every temporal slice y of x.
# Arguments
x: input tensor.
w: weight matrix.
b: optional bias vector.
dropout: wether to apply dropout (same dropout mask
for every temporal slice of the input).
input_dim: integer; optional dimensionality of the input.
output_dim: integer; optional dimensionality of the output.
timesteps: integer; optional number of timesteps.
training: training phase tensor or boolean.
# Returns
Output tensor.
"""
if not input_dim:
input_dim = K.shape(x)[2]
if not timesteps:
timesteps = K.shape(x)[1]
if not output_dim:
output_dim = K.int_shape(w)[1]
if dropout is not None and 0. < dropout < 1.:
# apply the same dropout pattern at every timestep
ones = K.ones_like(K.reshape(x[:, 0, :], (-1, input_dim)))
dropout_matrix = K.dropout(ones, dropout)
expanded_dropout_matrix = K.repeat(dropout_matrix, timesteps)
x = K.in_train_phase(x * expanded_dropout_matrix, x, training=training)
# collapse time dimension and batch dimension together
x = K.reshape(x, (-1, input_dim))
x = K.dot(x, w)
if b is not None:
x = K.bias_add(x, b)
# reshape to 3D tensor
if K.backend() == 'tensorflow':
x = K.reshape(x, K.stack([-1, timesteps, output_dim]))
x.set_shape([None, None, output_dim])
else:
x = K.reshape(x, (-1, timesteps, output_dim))
return x
def model_metadata(model, include_optimizer=True, require_config=True):
"""Returns a dictionary containing the model metadata."""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
from tensorflow.python.keras.optimizer_v2 import optimizer_v2 # pylint: disable=g-import-not-at-top
model_config = {'class_name': model.__class__.__name__}
try:
model_config['config'] = model.get_config()
except NotImplementedError as e:
if require_config:
raise e
metadata = dict(keras_version=str(keras_version),
backend=K.backend(),
model_config=model_config)
if model.optimizer and include_optimizer:
if isinstance(model.optimizer, optimizer_v1.TFOptimizer):
logging.warning(
'TensorFlow optimizers do not '
'make it possible to access '
'optimizer attributes or optimizer state '
'after instantiation. '
'As a result, we cannot save the optimizer '
'as part of the model save file. '
'You will have to compile your model again after loading it. '
'Prefer using a Keras optimizer instead '
'(see keras.io/optimizers).')
elif model._compile_was_called: # pylint: disable=protected-access
training_config = model._get_compile_args(user_metrics=False) # pylint: disable=protected-access
training_config.pop('optimizer', None) # Handled separately.
metadata['training_config'] = _serialize_nested_config(
training_config)
if isinstance(model.optimizer, optimizer_v2.RestoredOptimizer):
raise NotImplementedError(
'As of now, Optimizers loaded from SavedModel cannot be saved. '
'If you\'re calling `model.save` or `tf.keras.models.save_model`,'
' please set the `include_optimizer` option to `False`. For '
'`tf.saved_model.save`, delete the optimizer from the model.'
)
else:
optimizer_config = {
'class_name':
generic_utils.get_registered_name(
model.optimizer.__class__),
'config':
model.optimizer.get_config()
}
metadata['training_config']['optimizer_config'] = optimizer_config
return metadata
def lrn(x, alpha=1e-4, beta=0.75, k=2, n=5, name=None):
"""
Local response normalization layer.
Parameters:
----------
x : keras.backend tensor/variable/symbol
Input tensor/variable/symbol.
alpha : float, 1e-4
Variance scaling parameter alpha in the LRN expression.
beta : float, 0.75
Power parameter beta in the LRN expression.
k : float, 2
Parameter k in the LRN expression.
n : int, 5
Normalization window width in elements.
name : str, default None
Layer name.
Returns
-------
keras.backend tensor/variable/symbol
Resulted tensor/variable/symbol.
"""
if K.backend() == "mxnet":
from keras.backend.mxnet_backend import keras_mxnet_symbol, KerasSymbol
import mxnet as mx
@keras_mxnet_symbol
def gluon_lrn(x, alpha, beta, k, n):
if isinstance(x, KerasSymbol):
x = x.symbol
if isinstance(alpha, KerasSymbol):
alpha = alpha.symbol
if isinstance(beta, KerasSymbol):
beta = beta.symbol
if isinstance(k, KerasSymbol):
k = k.symbol
if isinstance(n, KerasSymbol):
n = n.symbol
return KerasSymbol(
mx.sym.LRN(data=x, alpha=alpha, beta=beta, knorm=k, nsize=n))
x = nn.Lambda(
lambda z: gluon_lrn(x=z, alpha=alpha, beta=beta, k=k, n=n))(x)
else:
import tensorflow as tf
x = nn.Lambda(lambda z: tf.nn.lrn(
input=z, depth_radius=n, bias=k, alpha=alpha, beta=beta, name=name)
)(x)
return x
def convert(in_path, out_path):
"""Convert any Keras model to the frugally-deep model format."""
assert K.backend() == "tensorflow"
assert K.floatx() == "float32"
assert K.image_data_format() == 'channels_last'
print('loading {}'.format(in_path))
model = load_model(in_path)
# Force creation of underlying functional model.
# see: https://github.com/fchollet/keras/issues/8136
# Loss and optimizer type do not matter, since to don't train the model.
model.compile(loss='mse', optimizer='sgd')
model = convert_sequential_to_model(model)
test_data = gen_test_data(model)
json_output = {}
json_output['architecture'] = json.loads(model.to_json())
json_output['image_data_format'] = K.image_data_format()
for depth in range(1, 3, 1):
json_output['conv2d_valid_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_conv2d_eval, 'valid')
json_output['conv2d_same_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_conv2d_eval, 'same')
json_output['separable_conv2d_valid_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_sep_conv2d_eval, 'valid')
json_output['separable_conv2d_same_offset_depth_' + str(depth)] =\
check_operation_offset(depth, offset_sep_conv2d_eval, 'same')
json_output['max_pooling_2d_valid_offset'] =\
check_operation_offset(1, conv2d_offset_max_pool_eval, 'valid')
json_output['max_pooling_2d_same_offset'] =\
check_operation_offset(1, conv2d_offset_max_pool_eval, 'same')
json_output['average_pooling_2d_valid_offset'] =\
check_operation_offset(1, conv2d_offset_average_pool_eval, 'valid')
json_output['average_pooling_2d_same_offset'] =\
check_operation_offset(1, conv2d_offset_average_pool_eval, 'same')
json_output['input_shapes'] = get_shapes(test_data['inputs'])
json_output['output_shapes'] = get_shapes(test_data['outputs'])
json_output['tests'] = [test_data]
json_output['trainable_params'] = get_all_weights(model)
print('writing {}'.format(out_path))
write_text_file(out_path, json.dumps(
json_output, allow_nan=False, indent=2, sort_keys=True))
def main():
"""Save famous example models in Keras format."""
if len(sys.argv) != 2:
print('usage: [output dir]')
sys.exit(1)
else:
assert K.backend() == "tensorflow"
assert K.floatx() == "float32"
assert K.image_data_format() == 'channels_last'
dir_path = sys.argv[1]
save_xception(dir_path, 'xception')
save_vgg16(dir_path, 'vgg16')
save_vgg19(dir_path, 'vgg19')
save_resnet50(dir_path, 'resnet50')
save_inceptionv3(dir_path, 'inceptionv3')
save_densenet201(dir_path, 'densenet201')
save_nasnetlarge(dir_path, 'nasnetlarge')
def main():
"""Convert any Keras model to the frugally-deep model format."""
usage = 'usage: [Keras model in HDF5 format] [image output directory]'
if len(sys.argv) != 3:
print(usage)
sys.exit(1)
else:
assert K.backend() == "tensorflow"
assert K.floatx() == "float32"
assert K.image_data_format() == 'channels_last'
in_path = sys.argv[1]
out_dir = sys.argv[2]
print('loading {}'.format(in_path))
K.set_learning_phase(1)
model = load_model(in_path)
model = convert_sequential_to_model(model)
process_layers(model, out_dir)
def save_weights_to_hdf5_group(f, layers):
"""Saves the weights of a list of layers to a HDF5 group.
Arguments:
f: HDF5 group.
layers: List of layer instances.
"""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
save_attributes_to_hdf5_group(
f, 'layer_names', [layer.name.encode('utf8') for layer in layers])
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['keras_version'] = str(keras_version).encode('utf8')
# On TPUs, modifying the graph between session.runs() triggers some expensive
# recompilation overhead. To avoid this, we build up the full set of tensors
# to save before fetching weights, thus only modifying the graph once.
layer_weights_dict = {}
for layer in layers:
layer_weights_dict[layer.name] = [
ops.convert_to_tensor(w) for w in layer.weights
]
for layer in layers:
g = f.create_group(layer.name)
symbolic_weights = layer_weights_dict[layer.name]
weight_values = K.batch_get_value(symbolic_weights)
weight_names = []
for i, (w, val) in enumerate(zip(symbolic_weights, weight_values)):
if hasattr(w, 'name') and w.name:
name = str(w.name)
else:
name = 'param_' + str(i)
weight_names.append(name.encode('utf8'))
save_attributes_to_hdf5_group(g, 'weight_names', weight_names)
for name, val in zip(weight_names, weight_values):
param_dset = g.create_dataset(name, val.shape, dtype=val.dtype)
if not val.shape:
# scalar
param_dset[()] = val
else:
param_dset[:] = val
def call(self, x, mask=None):
b, ch, r, c = x.shape
half_n = self.n // 2 # half the local region
input_sqr = K.square(x) # square the input
if K.backend() == 'theano':
# make an empty tensor with zero pads along channel dimension
zeros = T.alloc(0., b, ch + 2 * half_n, r, c)
# set the center to be the squared input
input_sqr = T.set_subtensor(zeros[:, half_n:half_n + ch, :, :],
input_sqr)
else:
input_sqr = tf.pad(input_sqr,
[[0, 0], [half_n, half_n], [0, 0], [0, 0]])
scale = self.k # offset for the scale
norm_alpha = self.alpha / self.n # normalized alpha
for i in range(self.n):
scale += norm_alpha * input_sqr[:, i:i + ch, :, :]
scale = scale**self.beta
x = x / scale
return x
def save_weights_to_hdf5_group(f, layers):
"""Saves the weights of a list of layers to a HDF5 group.
Arguments:
f: HDF5 group.
layers: List of layer instances.
"""
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
save_attributes_to_hdf5_group(
f, 'layer_names', [layer.name.encode('utf8') for layer in layers])
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['keras_version'] = str(keras_version).encode('utf8')
# On TPUs, modifying the graph between session.runs() triggers some expensive
# recompilation overhead. To avoid this, we build up the full set of tensors
# to save before fetching weights, thus only modifying the graph once.
layer_weights_dict = {}
for layer in layers:
layer_weights_dict[layer.name] = [ops.convert_to_tensor(w)
for w in layer.weights]
for layer in layers:
g = f.create_group(layer.name)
symbolic_weights = layer_weights_dict[layer.name]
weight_values = K.batch_get_value(symbolic_weights)
weight_names = []
for i, (w, val) in enumerate(zip(symbolic_weights, weight_values)):
if hasattr(w, 'name') and w.name:
name = str(w.name)
else:
name = 'param_' + str(i)
weight_names.append(name.encode('utf8'))
save_attributes_to_hdf5_group(g, 'weight_names', weight_names)
for name, val in zip(weight_names, weight_values):
param_dset = g.create_dataset(name, val.shape, dtype=val.dtype)
if not val.shape:
# scalar
param_dset[()] = val
else:
param_dset[:] = val
def save_model(model, filepath, overwrite=True, include_optimizer=True):
"""Saves a model to a HDF5 file.
The saved model contains:
- the model's configuration (topology)
- the model's weights
- the model's optimizer's state (if any)
Thus the saved model can be reinstantiated in
the exact same state, without any of the code
used for model definition or training.
Arguments:
model: Keras model instance to be saved.
filepath: One of the following:
- String, path where to save the model
- `h5py.File` object where to save the model
overwrite: Whether we should overwrite any existing
model at the target location, or instead
ask the user with a manual prompt.
include_optimizer: If True, save optimizer's state together.
Raises:
ImportError: if h5py is not available.
"""
if h5py is None:
raise ImportError('`save_model` requires h5py.')
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
# TODO(psv) Add warning when we save models that contain non-serializable
# entities like metrics added using `add_metric` and losses added using
# `add_loss.`
if not isinstance(filepath, h5py.File):
# If file exists and should not be overwritten.
if not overwrite and os.path.isfile(filepath):
proceed = ask_to_proceed_with_overwrite(filepath)
if not proceed:
return
f = h5py.File(filepath, mode='w')
opened_new_file = True
else:
f = filepath
opened_new_file = False
try:
f.attrs['keras_version'] = str(keras_version).encode('utf8')
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['model_config'] = json.dumps(
{
'class_name': model.__class__.__name__,
'config': model.get_config()
},
default=serialization.get_json_type).encode('utf8')
model_weights_group = f.create_group('model_weights')
model_layers = model.layers
save_weights_to_hdf5_group(model_weights_group, model_layers)
if include_optimizer and model.optimizer:
if isinstance(model.optimizer, optimizers.TFOptimizer):
logging.warning(
'TensorFlow optimizers do not '
'make it possible to access '
'optimizer attributes or optimizer state '
'after instantiation. '
'As a result, we cannot save the optimizer '
'as part of the model save file.'
'You will have to compile your model again after loading it. '
'Prefer using a Keras optimizer instead '
'(see keras.io/optimizers).')
else:
f.attrs['training_config'] = json.dumps(
{
'optimizer_config': {
'class_name': model.optimizer.__class__.__name__,
'config': model.optimizer.get_config()
},
'loss': model.loss,
'metrics': model._compile_metrics,
'weighted_metrics': model._compile_weighted_metrics,
'sample_weight_mode': model.sample_weight_mode,
'loss_weights': model.loss_weights,
},
default=serialization.get_json_type).encode('utf8')
# Save optimizer weights.
save_optimizer_weights_to_hdf5_group(f, model.optimizer)
f.flush()
finally:
if opened_new_file:
f.close()
def InceptionV3(include_top=True,
weights='imagenet',
input_tensor=None,
model_input=None,
pooling=None,
classes=1000, model_path=""):
"""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)
or 'imagenet' (pre-training on ImageNet).
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 weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
img_input = model_input
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`.
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 = model_path
model.load_weights(weights_path)
else:
weights_path = ""
elif (weights == "trained"):
weights_path = model_path
model.load_weights(weights_path)
return model
def DenseNet(model_input, depth=40, nb_dense_block=3, growth_rate=12, nb_filter=-1, nb_layers_per_block=-1,
bottleneck=False, reduction=0.0, dropout_rate=0.0, weight_decay=1e-4, subsample_initial_block=False,
include_top=True, weights=None, input_tensor=None,
classes=10, activation='softmax', model_path = ''):
'''Instantiate the DenseNet architecture,
optionally loading weights pre-trained
on CIFAR-10. 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 dimension ordering
convention used by the model is the one
specified in your Keras config file.
# Arguments
input_shape: optional shape tuple, only to be specified
if `include_top` is False (otherwise the input shape
has to be `(32, 32, 3)` (with `channels_last` dim ordering)
or `(3, 32, 32)` (with `channels_first` dim ordering).
It should have exactly 3 inputs channels,
and width and height should be no smaller than 8.
E.g. `(200, 200, 3)` would be one valid value.
depth: number or layers in the DenseNet
nb_dense_block: number of dense blocks to add to end (generally = 3)
growth_rate: number of filters to add per dense block
nb_filter: initial number of filters. -1 indicates initial
number of filters is 2 * growth_rate
nb_layers_per_block: number of layers in each dense block.
Can be a -1, positive integer or a list.
If -1, calculates nb_layer_per_block from the network depth.
If positive integer, a set number of layers per dense block.
If list, nb_layer is used as provided. Note that list size must
be (nb_dense_block + 1)
bottleneck: flag to add bottleneck blocks in between dense blocks
reduction: reduction factor of transition blocks.
Note : reduction value is inverted to compute compression.
dropout_rate: dropout rate
weight_decay: weight decay rate
subsample_initial_block: Set to True to subsample the initial convolution and
add a MaxPool2D before the dense blocks are added.
include_top: whether to include the fully-connected
layer at the top of the network.
weights: one of `None` (random initialization) or
'imagenet' (pre-training on ImageNet)..
input_tensor: optional Keras tensor (i.e. output of `layers.Input()`)
to use as image input for the model.
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.
activation: Type of activation at the top layer. Can be one of 'softmax' or 'sigmoid'.
Note that if sigmoid is used, classes must be 1.
# Returns
A Keras model instance.
'''
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 activation not in ['softmax', 'sigmoid']:
raise ValueError('activation must be one of "softmax" or "sigmoid"')
if activation == 'sigmoid' and classes != 1:
raise ValueError('sigmoid activation can only be used when classes = 1')
# Determine proper input shape
img_input = model_input
x = __create_dense_net(classes, img_input, include_top, depth, nb_dense_block,
growth_rate, nb_filter, nb_layers_per_block, bottleneck, reduction,
dropout_rate, weight_decay, subsample_initial_block, activation)
# Ensure that the model takes into account
# any potential predecessors of `input_tensor`.
inputs = img_input
# Create model.
model = Model(inputs, x, name='densenet')
# load weights
if weights == 'imagenet':
weights_loaded = False
if (depth == 121) and (nb_dense_block == 4) and (growth_rate == 32) and (nb_filter == 64) and \
(bottleneck is True) and (reduction == 0.5) and (dropout_rate == 0.0) and (subsample_initial_block):
weights_path = model_path
model.load_weights(weights_path)
weights_loaded = True
if weights_loaded:
if K.backend() == 'theano':
convert_all_kernels_in_model(model)
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 == "trained"):
weights_path = model_path
model.load_weights(weights_path)
return model
def NASNet(input_shape=None,
penultimate_filters=4032,
num_blocks=6,
stem_block_filters=96,
skip_reduction=True,
filter_multiplier=2,
include_top=True,
weights=None,
input_tensor=None,
pooling=None,
classes=1000,
default_size=None):
"""Instantiates a NASNet model.
Note that only TensorFlow is supported for now,
therefore it only works with the data format
`image_data_format='channels_last'` in your Keras config
at `~/.keras/keras.json`.
Arguments:
input_shape: Optional shape tuple, the input shape
is by default `(331, 331, 3)` for NASNetLarge and
`(224, 224, 3)` for NASNetMobile.
It should have exactly 3 inputs channels,
and width and height should be no smaller than 32.
E.g. `(224, 224, 3)` would be one valid value.
penultimate_filters: Number of filters in the penultimate layer.
NASNet models use the notation `NASNet (N @ P)`, where:
- N is the number of blocks
- P is the number of penultimate filters
num_blocks: Number of repeated blocks of the NASNet model.
NASNet models use the notation `NASNet (N @ P)`, where:
- N is the number of blocks
- P is the number of penultimate filters
stem_block_filters: Number of filters in the initial stem block
skip_reduction: Whether to skip the reduction step at the tail
end of the network. Set to `False` for CIFAR models.
filter_multiplier: Controls the width of the network.
- If `filter_multiplier` < 1.0, proportionally decreases the number
of filters in each layer.
- If `filter_multiplier` > 1.0, proportionally increases the number
of filters in each layer.
- If `filter_multiplier` = 1, default number of filters from the
paper are used at each layer.
include_top: Whether to include the fully-connected
layer at the top of the network.
weights: `None` (random initialization) or
`imagenet` (ImageNet weights)
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.
default_size: Specifies the default image size of the model
Returns:
A Keras model instance.
Raises:
ValueError: In case of invalid argument for `weights`,
invalid input shape or invalid `penultimate_filters` value.
RuntimeError: If attempting to run this model with a
backend that does not support separable convolutions.
"""
if K.backend() != 'tensorflow':
raise RuntimeError('Only Tensorflow backend is currently supported, '
'as other backends do not support '
'separable convolution.')
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 (isinstance(input_shape, tuple) and None in input_shape and
weights == 'imagenet'):
raise ValueError('When specifying the input shape of a NASNet'
' and loading `ImageNet` weights, '
'the input_shape argument must be static '
'(no None entries). Got: `input_shape=' +
str(input_shape) + '`.')
if default_size is None:
default_size = 331
# Determine proper input shape and default size.
input_shape = _obtain_input_shape(
input_shape,
default_size=default_size,
min_size=32,
data_format=K.image_data_format(),
require_flatten=False,
weights=weights)
if K.image_data_format() != 'channels_last':
logging.warning('The NASNet 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.')
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
if penultimate_filters % 24 != 0:
raise ValueError(
'For NASNet-A models, the value of `penultimate_filters` '
'needs to be divisible by 24. Current value: %d' % penultimate_filters)
channel_dim = 1 if K.image_data_format() == 'channels_first' else -1
filters = penultimate_filters // 24
if not skip_reduction:
x = Conv2D(
stem_block_filters, (3, 3),
strides=(2, 2),
padding='valid',
use_bias=False,
name='stem_conv1',
kernel_initializer='he_normal')(
img_input)
else:
x = Conv2D(
stem_block_filters, (3, 3),
strides=(1, 1),
padding='same',
use_bias=False,
name='stem_conv1',
kernel_initializer='he_normal')(
img_input)
x = BatchNormalization(
axis=channel_dim, momentum=0.9997, epsilon=1e-3, name='stem_bn1')(
x)
p = None
if not skip_reduction: # imagenet / mobile mode
x, p = _reduction_a_cell(
x, p, filters // (filter_multiplier**2), block_id='stem_1')
x, p = _reduction_a_cell(
x, p, filters // filter_multiplier, block_id='stem_2')
for i in range(num_blocks):
x, p = _normal_a_cell(x, p, filters, block_id='%d' % (i))
x, p0 = _reduction_a_cell(
x, p, filters * filter_multiplier, block_id='reduce_%d' % (num_blocks))
p = p0 if not skip_reduction else p
for i in range(num_blocks):
x, p = _normal_a_cell(
x, p, filters * filter_multiplier, block_id='%d' % (num_blocks + i + 1))
x, p0 = _reduction_a_cell(
x,
p,
filters * filter_multiplier**2,
block_id='reduce_%d' % (2 * num_blocks))
p = p0 if not skip_reduction else p
for i in range(num_blocks):
x, p = _normal_a_cell(
x,
p,
filters * filter_multiplier**2,
block_id='%d' % (2 * num_blocks + i + 1))
x = Activation('relu')(x)
if include_top:
x = GlobalAveragePooling2D()(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 = layer_utils.get_source_inputs(input_tensor)
else:
inputs = img_input
model = Model(inputs, x, name='NASNet')
# load weights
if weights == 'imagenet':
if default_size == 224: # mobile version
if include_top:
weight_path = NASNET_MOBILE_WEIGHT_PATH
model_name = 'nasnet_mobile.h5'
else:
weight_path = NASNET_MOBILE_WEIGHT_PATH_NO_TOP
model_name = 'nasnet_mobile_no_top.h5'
weights_file = get_file(model_name, weight_path, cache_subdir='models')
model.load_weights(weights_file)
elif default_size == 331: # large version
if include_top:
weight_path = NASNET_LARGE_WEIGHT_PATH
model_name = 'nasnet_large.h5'
else:
weight_path = NASNET_LARGE_WEIGHT_PATH_NO_TOP
model_name = 'nasnet_large_no_top.h5'
weights_file = get_file(model_name, weight_path, cache_subdir='models')
model.load_weights(weights_file)
else:
raise ValueError('ImageNet weights can only be loaded with NASNetLarge'
' or NASNetMobile')
elif weights is not None:
model.load_weights(weights)
if old_data_format:
K.set_image_data_format(old_data_format)
return model
def save_model(model, filepath, overwrite=True, include_optimizer=True):
"""Saves a model to a HDF5 file.
The saved model contains:
- the model's configuration (topology)
- the model's weights
- the model's optimizer's state (if any)
Thus the saved model can be reinstantiated in
the exact same state, without any of the code
used for model definition or training.
Arguments:
model: Keras model instance to be saved.
filepath: One of the following:
- String, path where to save the model
- `h5py.File` object where to save the model
overwrite: Whether we should overwrite any existing
model at the target location, or instead
ask the user with a manual prompt.
include_optimizer: If True, save optimizer's state together.
Raises:
ImportError: if h5py is not available.
"""
if h5py is None:
raise ImportError('`save_model` requires h5py.')
from tensorflow.python.keras import __version__ as keras_version # pylint: disable=g-import-not-at-top
if not isinstance(filepath, h5py.File):
# If file exists and should not be overwritten.
if not overwrite and os.path.isfile(filepath):
proceed = ask_to_proceed_with_overwrite(filepath)
if not proceed:
return
f = h5py.File(filepath, mode='w')
opened_new_file = True
else:
f = filepath
opened_new_file = False
try:
f.attrs['keras_version'] = str(keras_version).encode('utf8')
f.attrs['backend'] = K.backend().encode('utf8')
f.attrs['model_config'] = json.dumps(
{
'class_name': model.__class__.__name__,
'config': model.get_config()
},
default=serialization.get_json_type).encode('utf8')
model_weights_group = f.create_group('model_weights')
model_layers = model.layers
save_weights_to_hdf5_group(model_weights_group, model_layers)
if include_optimizer and hasattr(model, 'optimizer'):
if isinstance(model.optimizer, optimizers.TFOptimizer):
logging.warning(
'TensorFlow optimizers do not '
'make it possible to access '
'optimizer attributes or optimizer state '
'after instantiation. '
'As a result, we cannot save the optimizer '
'as part of the model save file.'
'You will have to compile your model again after loading it. '
'Prefer using a Keras optimizer instead '
'(see keras.io/optimizers).')
else:
f.attrs['training_config'] = json.dumps(
{
'optimizer_config': {
'class_name': model.optimizer.__class__.__name__,
'config': model.optimizer.get_config()
},
'loss': model.loss,
'metrics': model.metrics,
'sample_weight_mode': model.sample_weight_mode,
'loss_weights': model.loss_weights,
},
default=serialization.get_json_type).encode('utf8')
# Save optimizer weights.
symbolic_weights = getattr(model.optimizer, 'weights')
if symbolic_weights:
optimizer_weights_group = f.create_group(
'optimizer_weights')
weight_values = K.batch_get_value(symbolic_weights)
weight_names = []
for w, val in zip(symbolic_weights, weight_values):
name = str(w.name)
weight_names.append(name.encode('utf8'))
optimizer_weights_group.attrs[
'weight_names'] = weight_names
for name, val in zip(weight_names, weight_values):
param_dset = optimizer_weights_group.create_dataset(
name, val.shape, dtype=val.dtype)
if not val.shape:
# scalar
param_dset[()] = val
else:
param_dset[:] = val
f.flush()
finally:
if opened_new_file:
f.close()
def __init__(self,
confidence_thresh=0.01,
iou_threshold=0.45,
top_k=200,
nms_max_output_size=400,
coords='centroids',
normalize_coords=True,
img_height=None,
img_width=None,
**kwargs):
'''
All default argument values follow the Caffe implementation.
Arguments:
confidence_thresh (float, optional): A float in [0,1), the minimum classification confidence in a specific
positive class in order to be considered for the non-maximum suppression stage for the respective class.
A lower value will result in a larger part of the selection process being done by the non-maximum suppression
stage, while a larger value will result in a larger part of the selection process happening in the confidence
thresholding stage.
iou_threshold (float, optional): A float in [0,1]. All boxes with a Jaccard similarity of greater than `iou_threshold`
with a locally maximal box will be removed from the set of predictions for a given class, where 'maximal' refers
to the box score.
top_k (int, optional): The number of highest scoring predictions to be kept for each batch item after the
non-maximum suppression stage.
nms_max_output_size (int, optional): The maximum number of predictions that will be left after performing non-maximum
suppression.
coords (str, optional): The box coordinate format that the model outputs. Must be 'centroids'
i.e. the format `(cx, cy, w, h)` (box center coordinates, width, and height). Other coordinate formats are
currently not supported.
normalize_coords (bool, optional): Set to `True` if the model outputs relative coordinates (i.e. coordinates in [0,1])
and you wish to transform these relative coordinates back to absolute coordinates. If the model outputs
relative coordinates, but you do not want to convert them back to absolute coordinates, set this to `False`.
Do not set this to `True` if the model already outputs absolute coordinates, as that would result in incorrect
coordinates. Requires `img_height` and `img_width` if set to `True`.
img_height (int, optional): The height of the input images. Only needed if `normalize_coords` is `True`.
img_width (int, optional): The width of the input images. Only needed if `normalize_coords` is `True`.
'''
if K.backend() != 'tensorflow':
raise TypeError(
"This layer only supports TensorFlow at the moment, but you are using the {} backend."
.format(K.backend()))
if normalize_coords and ((img_height is None) or (img_width is None)):
raise ValueError(
"If relative box coordinates are supposed to be converted to absolute coordinates, the decoder needs the image size in order to decode the predictions, but `img_height == {}` and `img_width == {}`"
.format(img_height, img_width))
if coords != 'centroids':
raise ValueError(
"The DetectionOutput layer currently only supports the 'centroids' coordinate format."
)
# We need these members for the config.
self.confidence_thresh = confidence_thresh
self.iou_threshold = iou_threshold
self.top_k = top_k
self.normalize_coords = normalize_coords
self.img_height = img_height
self.img_width = img_width
self.coords = coords
self.nms_max_output_size = nms_max_output_size
# We need these members for TensorFlow.
self.tf_confidence_thresh = tf.constant(self.confidence_thresh,
name='confidence_thresh')
self.tf_iou_threshold = tf.constant(self.iou_threshold,
name='iou_threshold')
self.tf_top_k = tf.constant(self.top_k, name='top_k')
self.tf_normalize_coords = tf.constant(self.normalize_coords,
name='normalize_coords')
self.tf_img_height = tf.constant(self.img_height,
dtype=tf.float32,
name='img_height')
self.tf_img_width = tf.constant(self.img_width,
dtype=tf.float32,
name='img_width')
self.tf_nms_max_output_size = tf.constant(self.nms_max_output_size,
name='nms_max_output_size')
super(DecodeDetectionsFast, self).__init__(**kwargs)
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=128,
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 = ZeroPadding2D(padding=(3, 3), name='conv1_pad')(img_input)
x = Conv2D(64, (7, 7),
strides=(2, 2),
padding='valid',
name='conv1',
kernel_regularizer=l2(0.0001))(x)
x = BatchNormalization(axis=bn_axis, name='bn_conv1')(x)
x = Activation('relu', name="act_conv1")(x)
x = MaxPooling2D((3, 3), strides=(2, 2), 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')
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')
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
