TensorFlow：包含卷积层类及其功能别名
2018-08-10 10:54 更新
# 版权所有2015 The TensorFlow作者.版权所有
# 根据Apache许可证2.0版(“许可证”)获得许可;
# 除非符合许可,否则您不得使用此文件.
# 您可以在以下位置获取许可证副本：http://www.apache.org/licenses/LICENSE-2.0
# 除非适用法律要求或书面同意,否则软件
# 根据许可证分发的“按现状”分发,
# 没有明示或暗示的任何形式的保证或条件.
# 有关管理权限的特定语言,请参阅许可证
# 许可证下的限制.
# =============================================================================
# pylint: disable=unused-import,g-bad-import-order
"""包含卷积层类及其功能别名"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from tensorflow.python.framework import tensor_shape
from tensorflow.python.layers import base
from tensorflow.python.layers import utils
from tensorflow.python.ops import array_ops
from tensorflow.python.ops import init_ops
from tensorflow.python.ops import nn
from tensorflow.python.ops import nn_ops
from tensorflow.python.util.tf_export import tf_export
class _Conv(base.Layer):
"""Abstract nD convolution layer (private, used as implementation base).
This layer creates a convolution kernel that is convolved
(actually cross-correlated) with the layer input to produce a tensor of
outputs. If `use_bias` is True (and a `bias_initializer` is provided),
a bias vector is created and added to the outputs. Finally, if
`activation` is not `None`, it is applied to the outputs as well.
Arguments:
rank: An integer, the rank of the convolution, e.g. "2" for 2D convolution.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of n integers, specifying the
length of the convolution window.
strides: An integer or tuple/list of n integers,
specifying the stride length of the convolution.
Specifying any stride value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, ..., channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, ...)`.
dilation_rate: An integer or tuple/list of n integers, specifying
the dilation rate to use for dilated convolution.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any `strides` value != 1.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self, rank,
filters,
kernel_size,
strides=1,
padding='valid',
data_format='channels_last',
dilation_rate=1,
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(_Conv, self).__init__(trainable=trainable, name=name,
activity_regularizer=activity_regularizer,
**kwargs)
self.rank = rank
self.filters = filters
self.kernel_size = utils.normalize_tuple(kernel_size, rank, 'kernel_size')
self.strides = utils.normalize_tuple(strides, rank, 'strides')
self.padding = utils.normalize_padding(padding)
self.data_format = utils.normalize_data_format(data_format)
self.dilation_rate = utils.normalize_tuple(
dilation_rate, rank, 'dilation_rate')
self.activation = activation
self.use_bias = use_bias
self.kernel_initializer = kernel_initializer
self.bias_initializer = bias_initializer
self.kernel_regularizer = kernel_regularizer
self.bias_regularizer = bias_regularizer
self.kernel_constraint = kernel_constraint
self.bias_constraint = bias_constraint
self.input_spec = base.InputSpec(ndim=self.rank + 2)
def build(self, input_shape):
input_shape = tensor_shape.TensorShape(input_shape)
if self.data_format == 'channels_first':
channel_axis = 1
else:
channel_axis = -1
if input_shape[channel_axis].value is None:
raise ValueError('The channel dimension of the inputs '
'should be defined. Found `None`.')
input_dim = input_shape[channel_axis].value
kernel_shape = self.kernel_size + (input_dim, self.filters)
self.kernel = self.add_variable(name='kernel',
shape=kernel_shape,
initializer=self.kernel_initializer,
regularizer=self.kernel_regularizer,
constraint=self.kernel_constraint,
trainable=True,
dtype=self.dtype)
if self.use_bias:
self.bias = self.add_variable(name='bias',
shape=(self.filters,),
initializer=self.bias_initializer,
regularizer=self.bias_regularizer,
constraint=self.bias_constraint,
trainable=True,
dtype=self.dtype)
else:
self.bias = None
self.input_spec = base.InputSpec(ndim=self.rank + 2,
axes={channel_axis: input_dim})
self._convolution_op = nn_ops.Convolution(
input_shape,
filter_shape=self.kernel.get_shape(),
dilation_rate=self.dilation_rate,
strides=self.strides,
padding=self.padding.upper(),
data_format=utils.convert_data_format(self.data_format,
self.rank + 2))
self.built = True
def call(self, inputs):
outputs = self._convolution_op(inputs, self.kernel)
if self.use_bias:
if self.data_format == 'channels_first':
if self.rank == 1:
# nn.bias_add does not accept a 1D input tensor.
bias = array_ops.reshape(self.bias, (1, self.filters, 1))
outputs += bias
if self.rank == 2:
outputs = nn.bias_add(outputs, self.bias, data_format='NCHW')
if self.rank == 3:
# As of Mar 2017, direct addition is significantly slower than
# bias_add when computing gradients. To use bias_add, we collapse Z
# and Y into a single dimension to obtain a 4D input tensor.
outputs_shape = outputs.shape.as_list()
if outputs_shape[0] is None:
outputs_shape[0] = -1
outputs_4d = array_ops.reshape(outputs,
[outputs_shape[0], outputs_shape[1],
outputs_shape[2] * outputs_shape[3],
outputs_shape[4]])
outputs_4d = nn.bias_add(outputs_4d, self.bias, data_format='NCHW')
outputs = array_ops.reshape(outputs_4d, outputs_shape)
else:
outputs = nn.bias_add(outputs, self.bias, data_format='NHWC')
if self.activation is not None:
return self.activation(outputs)
return outputs
def compute_output_shape(self, input_shape):
input_shape = tensor_shape.TensorShape(input_shape).as_list()
if self.data_format == 'channels_last':
space = input_shape[1:-1]
new_space = []
for i in range(len(space)):
new_dim = utils.conv_output_length(
space[i],
self.kernel_size[i],
padding=self.padding,
stride=self.strides[i],
dilation=self.dilation_rate[i])
new_space.append(new_dim)
return tensor_shape.TensorShape([input_shape[0]] + new_space +
[self.filters])
else:
space = input_shape[2:]
new_space = []
for i in range(len(space)):
new_dim = utils.conv_output_length(
space[i],
self.kernel_size[i],
padding=self.padding,
stride=self.strides[i],
dilation=self.dilation_rate[i])
new_space.append(new_dim)
return tensor_shape.TensorShape([input_shape[0], self.filters] +
new_space)
@tf_export('layers.Conv1D')
class Conv1D(_Conv):
"""1D convolution layer (e.g. temporal convolution).
This layer creates a convolution kernel that is convolved
(actually cross-correlated) with the layer input to produce a tensor of
outputs. If `use_bias` is True (and a `bias_initializer` is provided),
a bias vector is created and added to the outputs. Finally, if
`activation` is not `None`, it is applied to the outputs as well.
Arguments:
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of a single integer, specifying the
length of the 1D convolution window.
strides: An integer or tuple/list of a single integer,
specifying the stride length of the convolution.
Specifying any stride value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, length, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, length)`.
dilation_rate: An integer or tuple/list of a single integer, specifying
the dilation rate to use for dilated convolution.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any `strides` value != 1.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self, filters,
kernel_size,
strides=1,
padding='valid',
data_format='channels_last',
dilation_rate=1,
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(Convolution1D, self).__init__(
rank=1,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name, **kwargs)
@tf_export('layers.conv1d')
def conv1d(inputs,
filters,
kernel_size,
strides=1,
padding='valid',
data_format='channels_last',
dilation_rate=1,
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None):
"""Functional interface for 1D convolution layer (e.g. temporal convolution).
This layer creates a convolution kernel that is convolved
(actually cross-correlated) with the layer input to produce a tensor of
outputs. If `use_bias` is True (and a `bias_initializer` is provided),
a bias vector is created and added to the outputs. Finally, if
`activation` is not `None`, it is applied to the outputs as well.
Arguments:
inputs: Tensor input.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of a single integer, specifying the
length of the 1D convolution window.
strides: An integer or tuple/list of a single integer,
specifying the stride length of the convolution.
Specifying any stride value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, length, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, length)`.
dilation_rate: An integer or tuple/list of a single integer, specifying
the dilation rate to use for dilated convolution.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any `strides` value != 1.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
Output tensor.
Raises:
ValueError: if eager execution is enabled.
"""
layer = Conv1D(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
dtype=inputs.dtype.base_dtype,
_reuse=reuse,
_scope=name)
return layer.apply(inputs)
@tf_export('layers.Conv2D')
class Conv2D(_Conv):
"""2D convolution layer (e.g. spatial convolution over images).
This layer creates a convolution kernel that is convolved
(actually cross-correlated) with the layer input to produce a tensor of
outputs. If `use_bias` is True (and a `bias_initializer` is provided),
a bias vector is created and added to the outputs. Finally, if
`activation` is not `None`, it is applied to the outputs as well.
Arguments:
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of 2 integers, specifying the
height and width of the 2D convolution window.
Can be a single integer to specify the same value for
all spatial dimensions.
strides: An integer or tuple/list of 2 integers,
specifying the strides of the convolution along the height and width.
Can be a single integer to specify the same value for
all spatial dimensions.
Specifying any stride value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, height, width, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, height, width)`.
dilation_rate: An integer or tuple/list of 2 integers, specifying
the dilation rate to use for dilated convolution.
Can be a single integer to specify the same value for
all spatial dimensions.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self, filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1),
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(Conv2D, self).__init__(
rank=2,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name, **kwargs)
@tf_export('layers.conv2d')
def conv2d(inputs,
filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1),
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None):
"""Functional interface for the 2D convolution layer.
This layer creates a convolution kernel that is convolved
(actually cross-correlated) with the layer input to produce a tensor of
outputs. If `use_bias` is True (and a `bias_initializer` is provided),
a bias vector is created and added to the outputs. Finally, if
`activation` is not `None`, it is applied to the outputs as well.
Arguments:
inputs: Tensor input.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of 2 integers, specifying the
height and width of the 2D convolution window.
Can be a single integer to specify the same value for
all spatial dimensions.
strides: An integer or tuple/list of 2 integers,
specifying the strides of the convolution along the height and width.
Can be a single integer to specify the same value for
all spatial dimensions.
Specifying any stride value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, height, width, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, height, width)`.
dilation_rate: An integer or tuple/list of 2 integers, specifying
the dilation rate to use for dilated convolution.
Can be a single integer to specify the same value for
all spatial dimensions.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
Output tensor.
Raises:
ValueError: if eager execution is enabled.
"""
layer = Conv2D(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
dtype=inputs.dtype.base_dtype,
_reuse=reuse,
_scope=name)
return layer.apply(inputs)
@tf_export('layers.Conv3D')
class Conv3D(_Conv):
"""3D convolution layer (e.g. spatial convolution over volumes).
This layer creates a convolution kernel that is convolved
(actually cross-correlated) with the layer input to produce a tensor of
outputs. If `use_bias` is True (and a `bias_initializer` is provided),
a bias vector is created and added to the outputs. Finally, if
`activation` is not `None`, it is applied to the outputs as well.
Arguments:
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of 3 integers, specifying the
depth, height and width of the 3D convolution window.
Can be a single integer to specify the same value for
all spatial dimensions.
strides: An integer or tuple/list of 3 integers,
specifying the strides of the convolution along the depth,
height and width.
Can be a single integer to specify the same value for
all spatial dimensions.
Specifying any stride value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, depth, height, width, channels)` while `channels_first`
corresponds to inputs with shape
`(batch, channels, depth, height, width)`.
dilation_rate: An integer or tuple/list of 3 integers, specifying
the dilation rate to use for dilated convolution.
Can be a single integer to specify the same value for
all spatial dimensions.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self, filters,
kernel_size,
strides=(1, 1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1, 1),
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(Conv3D, self).__init__(
rank=3,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name, **kwargs)
@tf_export('layers.conv3d')
def conv3d(inputs,
filters,
kernel_size,
strides=(1, 1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1, 1),
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None):
"""Functional interface for the 3D convolution layer.
This layer creates a convolution kernel that is convolved
(actually cross-correlated) with the layer input to produce a tensor of
outputs. If `use_bias` is True (and a `bias_initializer` is provided),
a bias vector is created and added to the outputs. Finally, if
`activation` is not `None`, it is applied to the outputs as well.
Arguments:
inputs: Tensor input.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of 3 integers, specifying the
depth, height and width of the 3D convolution window.
Can be a single integer to specify the same value for
all spatial dimensions.
strides: An integer or tuple/list of 3 integers,
specifying the strides of the convolution along the depth,
height and width.
Can be a single integer to specify the same value for
all spatial dimensions.
Specifying any stride value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, depth, height, width, channels)` while `channels_first`
corresponds to inputs with shape
`(batch, channels, depth, height, width)`.
dilation_rate: An integer or tuple/list of 3 integers, specifying
the dilation rate to use for dilated convolution.
Can be a single integer to specify the same value for
all spatial dimensions.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
Output tensor.
Raises:
ValueError: if eager execution is enabled.
"""
layer = Conv3D(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
dtype=inputs.dtype.base_dtype,
_reuse=reuse,
_scope=name)
return layer.apply(inputs)
class _SeparableConv(_Conv):
"""Abstract base layer for separable nD convolution.
This layer performs a depthwise convolution that acts separately on
channels, followed by a pointwise convolution that mixes channels.
If `use_bias` is True and a bias initializer is provided,
it adds a bias vector to the output.
It then optionally applies an activation function to produce the final output.
Arguments:
rank: An integer, the rank of the convolution, e.g. "2" for 2D convolution.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A tuple or list of integers specifying the spatial
dimensions of the filters. Can be a single integer to specify the same
value for all spatial dimensions.
strides: A tuple or list of integers specifying the strides
of the convolution. Can be a single integer to specify the same value for
all spatial dimensions.
Specifying any `stride` value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, ..., channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, ...)`.
dilation_rate: An integer or tuple/list of 2 integers, specifying
the dilation rate to use for dilated convolution.
Can be a single integer to specify the same value for
all spatial dimensions.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
depth_multiplier: The number of depthwise convolution output channels for
each input channel. The total number of depthwise convolution output
channels will be equal to `num_filters_in * depth_multiplier`.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
depthwise_initializer: An initializer for the depthwise convolution kernel.
pointwise_initializer: An initializer for the pointwise convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
depthwise_regularizer: Optional regularizer for the depthwise
convolution kernel.
pointwise_regularizer: Optional regularizer for the pointwise
convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
depthwise_constraint: Optional projection function to be applied to the
depthwise kernel after being updated by an `Optimizer` (e.g. used for
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
pointwise_constraint: Optional projection function to be applied to the
pointwise kernel after being updated by an `Optimizer`.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self,
rank,
filters,
kernel_size,
strides=1,
padding='valid',
data_format='channels_last',
dilation_rate=1,
depth_multiplier=1,
activation=None,
use_bias=True,
depthwise_initializer=None,
pointwise_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
depthwise_regularizer=None,
pointwise_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
depthwise_constraint=None,
pointwise_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(_SeparableConv, self).__init__(
rank=rank,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
activation=activation,
use_bias=use_bias,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
**kwargs)
self.depth_multiplier = depth_multiplier
self.depthwise_initializer = depthwise_initializer
self.pointwise_initializer = pointwise_initializer
self.depthwise_regularizer = depthwise_regularizer
self.pointwise_regularizer = pointwise_regularizer
self.depthwise_constraint = depthwise_constraint
self.pointwise_constraint = pointwise_constraint
def build(self, input_shape):
input_shape = tensor_shape.TensorShape(input_shape)
if self.data_format == 'channels_first':
channel_axis = 1
else:
channel_axis = -1
if input_shape[channel_axis].value is None:
raise ValueError('The channel dimension of the inputs '
'should be defined. Found `None`.')
input_dim = input_shape[channel_axis].value
self.input_spec = base.InputSpec(ndim=self.rank + 2,
axes={channel_axis: input_dim})
depthwise_kernel_shape = self.kernel_size + (input_dim,
self.depth_multiplier)
pointwise_kernel_shape = (
1,) * self.rank + (self.depth_multiplier * input_dim, self.filters)
self.depthwise_kernel = self.add_variable(
name='depthwise_kernel',
shape=depthwise_kernel_shape,
initializer=self.depthwise_initializer,
regularizer=self.depthwise_regularizer,
constraint=self.depthwise_constraint,
trainable=True,
dtype=self.dtype)
self.pointwise_kernel = self.add_variable(
name='pointwise_kernel',
shape=pointwise_kernel_shape,
initializer=self.pointwise_initializer,
regularizer=self.pointwise_regularizer,
constraint=self.pointwise_constraint,
trainable=True,
dtype=self.dtype)
if self.use_bias:
self.bias = self.add_variable(name='bias',
shape=(self.filters,),
initializer=self.bias_initializer,
regularizer=self.bias_regularizer,
constraint=self.bias_constraint,
trainable=True,
dtype=self.dtype)
else:
self.bias = None
self.built = True
def call(self, inputs):
raise NotImplementedError
@tf_export('layers.SeparableConv1D')
class SeparableConv1D(_SeparableConv):
"""Depthwise separable 1D convolution.
This layer performs a depthwise convolution that acts separately on
channels, followed by a pointwise convolution that mixes channels.
If `use_bias` is True and a bias initializer is provided,
it adds a bias vector to the output.
It then optionally applies an activation function to produce the final output.
Arguments:
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A single integer specifying the spatial
dimensions of the filters.
strides: A single integer specifying the strides
of the convolution.
Specifying any `stride` value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, length, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, length)`.
dilation_rate: A single integer, specifying
the dilation rate to use for dilated convolution.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
depth_multiplier: The number of depthwise convolution output channels for
each input channel. The total number of depthwise convolution output
channels will be equal to `num_filters_in * depth_multiplier`.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
depthwise_initializer: An initializer for the depthwise convolution kernel.
pointwise_initializer: An initializer for the pointwise convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
depthwise_regularizer: Optional regularizer for the depthwise
convolution kernel.
pointwise_regularizer: Optional regularizer for the pointwise
convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
depthwise_constraint: Optional projection function to be applied to the
depthwise kernel after being updated by an `Optimizer` (e.g. used for
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
pointwise_constraint: Optional projection function to be applied to the
pointwise kernel after being updated by an `Optimizer`.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self, filters,
kernel_size,
strides=1,
padding='valid',
data_format='channels_last',
dilation_rate=1,
depth_multiplier=1,
activation=None,
use_bias=True,
depthwise_initializer=None,
pointwise_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
depthwise_regularizer=None,
pointwise_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
depthwise_constraint=None,
pointwise_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(SeparableConv1D, self).__init__(
rank=1,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
depth_multiplier=depth_multiplier,
activation=activation,
use_bias=use_bias,
depthwise_initializer=depthwise_initializer,
pointwise_initializer=pointwise_initializer,
bias_initializer=bias_initializer,
depthwise_regularizer=depthwise_regularizer,
pointwise_regularizer=pointwise_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
depthwise_constraint=depthwise_constraint,
pointwise_constraint=pointwise_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
**kwargs)
def call(self, inputs):
if self.data_format == 'channels_last':
strides = (1,) + self.strides * 2 + (1,)
spatial_start_dim = 1
else:
strides = (1, 1) + self.strides * 2
spatial_start_dim = 2
# Explicitly broadcast inputs and kernels to 4D.
# TODO(fchollet): refactor when a native separable_conv1d op is available.
inputs = array_ops.expand_dims(inputs, spatial_start_dim)
depthwise_kernel = array_ops.expand_dims(self.depthwise_kernel, 0)
pointwise_kernel = array_ops.expand_dims(self.pointwise_kernel, 0)
dilation_rate = (1,) + self.dilation_rate
outputs = nn.separable_conv2d(
inputs,
depthwise_kernel,
pointwise_kernel,
strides=strides,
padding=self.padding.upper(),
rate=dilation_rate,
data_format=utils.convert_data_format(self.data_format, ndim=4))
if self.use_bias:
outputs = nn.bias_add(
outputs,
self.bias,
data_format=utils.convert_data_format(self.data_format, ndim=4))
outputs = array_ops.squeeze(outputs, [spatial_start_dim])
if self.activation is not None:
return self.activation(outputs)
return outputs
@tf_export('layers.SeparableConv2D')
class SeparableConv2D(_SeparableConv):
"""Depthwise separable 2D convolution.
This layer performs a depthwise convolution that acts separately on
channels, followed by a pointwise convolution that mixes channels.
If `use_bias` is True and a bias initializer is provided,
it adds a bias vector to the output.
It then optionally applies an activation function to produce the final output.
Arguments:
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A tuple or list of 2 integers specifying the spatial
dimensions of the filters. Can be a single integer to specify the same
value for all spatial dimensions.
strides: A tuple or list of 2 positive integers specifying the strides
of the convolution. Can be a single integer to specify the same value for
all spatial dimensions.
Specifying any `stride` value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, height, width, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, height, width)`.
dilation_rate: An integer or tuple/list of 2 integers, specifying
the dilation rate to use for dilated convolution.
Can be a single integer to specify the same value for
all spatial dimensions.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
depth_multiplier: The number of depthwise convolution output channels for
each input channel. The total number of depthwise convolution output
channels will be equal to `num_filters_in * depth_multiplier`.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
depthwise_initializer: An initializer for the depthwise convolution kernel.
pointwise_initializer: An initializer for the pointwise convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
depthwise_regularizer: Optional regularizer for the depthwise
convolution kernel.
pointwise_regularizer: Optional regularizer for the pointwise
convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
depthwise_constraint: Optional projection function to be applied to the
depthwise kernel after being updated by an `Optimizer` (e.g. used for
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
pointwise_constraint: Optional projection function to be applied to the
pointwise kernel after being updated by an `Optimizer`.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self, filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1),
depth_multiplier=1,
activation=None,
use_bias=True,
depthwise_initializer=None,
pointwise_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
depthwise_regularizer=None,
pointwise_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
depthwise_constraint=None,
pointwise_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(SeparableConv2D, self).__init__(
rank=2,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
depth_multiplier=depth_multiplier,
activation=activation,
use_bias=use_bias,
depthwise_initializer=depthwise_initializer,
pointwise_initializer=pointwise_initializer,
bias_initializer=bias_initializer,
depthwise_regularizer=depthwise_regularizer,
pointwise_regularizer=pointwise_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
depthwise_constraint=depthwise_constraint,
pointwise_constraint=pointwise_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
**kwargs)
def call(self, inputs):
# Apply the actual ops.
if self.data_format == 'channels_last':
strides = (1,) + self.strides + (1,)
else:
strides = (1, 1) + self.strides
outputs = nn.separable_conv2d(
inputs,
self.depthwise_kernel,
self.pointwise_kernel,
strides=strides,
padding=self.padding.upper(),
rate=self.dilation_rate,
data_format=utils.convert_data_format(self.data_format, ndim=4))
if self.use_bias:
outputs = nn.bias_add(
outputs,
self.bias,
data_format=utils.convert_data_format(self.data_format, ndim=4))
if self.activation is not None:
return self.activation(outputs)
return outputs
@tf_export('layers.separable_conv1d')
def separable_conv1d(inputs,
filters,
kernel_size,
strides=1,
padding='valid',
data_format='channels_last',
dilation_rate=1,
depth_multiplier=1,
activation=None,
use_bias=True,
depthwise_initializer=None,
pointwise_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
depthwise_regularizer=None,
pointwise_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
depthwise_constraint=None,
pointwise_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None):
"""Functional interface for the depthwise separable 1D convolution layer.
This layer performs a depthwise convolution that acts separately on
channels, followed by a pointwise convolution that mixes channels.
If `use_bias` is True and a bias initializer is provided,
it adds a bias vector to the output.
It then optionally applies an activation function to produce the final output.
Arguments:
inputs: Input tensor.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A single integer specifying the spatial
dimensions of the filters.
strides: A single integer specifying the strides
of the convolution.
Specifying any `stride` value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, length, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, length)`.
dilation_rate: A single integer, specifying
the dilation rate to use for dilated convolution.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
depth_multiplier: The number of depthwise convolution output channels for
each input channel. The total number of depthwise convolution output
channels will be equal to `num_filters_in * depth_multiplier`.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
depthwise_initializer: An initializer for the depthwise convolution kernel.
pointwise_initializer: An initializer for the pointwise convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
depthwise_regularizer: Optional regularizer for the depthwise
convolution kernel.
pointwise_regularizer: Optional regularizer for the pointwise
convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
depthwise_constraint: Optional projection function to be applied to the
depthwise kernel after being updated by an `Optimizer` (e.g. used for
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
pointwise_constraint: Optional projection function to be applied to the
pointwise kernel after being updated by an `Optimizer`.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
Output tensor.
Raises:
ValueError: if eager execution is enabled.
"""
layer = SeparableConv1D(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
depth_multiplier=depth_multiplier,
activation=activation,
use_bias=use_bias,
depthwise_initializer=depthwise_initializer,
pointwise_initializer=pointwise_initializer,
bias_initializer=bias_initializer,
depthwise_regularizer=depthwise_regularizer,
pointwise_regularizer=pointwise_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
depthwise_constraint=depthwise_constraint,
pointwise_constraint=pointwise_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
_reuse=reuse,
_scope=name)
return layer.apply(inputs)
@tf_export('layers.separable_conv2d')
def separable_conv2d(inputs,
filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
dilation_rate=(1, 1),
depth_multiplier=1,
activation=None,
use_bias=True,
depthwise_initializer=None,
pointwise_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
depthwise_regularizer=None,
pointwise_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
depthwise_constraint=None,
pointwise_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None):
"""Functional interface for the depthwise separable 2D convolution layer.
This layer performs a depthwise convolution that acts separately on
channels, followed by a pointwise convolution that mixes channels.
If `use_bias` is True and a bias initializer is provided,
it adds a bias vector to the output.
It then optionally applies an activation function to produce the final output.
Arguments:
inputs: Input tensor.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A tuple or list of 2 integers specifying the spatial
dimensions of the filters. Can be a single integer to specify the same
value for all spatial dimensions.
strides: A tuple or list of 2 positive integers specifying the strides
of the convolution. Can be a single integer to specify the same value for
all spatial dimensions.
Specifying any `stride` value != 1 is incompatible with specifying
any `dilation_rate` value != 1.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, height, width, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, height, width)`.
dilation_rate: An integer or tuple/list of 2 integers, specifying
the dilation rate to use for dilated convolution.
Can be a single integer to specify the same value for
all spatial dimensions.
Currently, specifying any `dilation_rate` value != 1 is
incompatible with specifying any stride value != 1.
depth_multiplier: The number of depthwise convolution output channels for
each input channel. The total number of depthwise convolution output
channels will be equal to `num_filters_in * depth_multiplier`.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
depthwise_initializer: An initializer for the depthwise convolution kernel.
pointwise_initializer: An initializer for the pointwise convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
depthwise_regularizer: Optional regularizer for the depthwise
convolution kernel.
pointwise_regularizer: Optional regularizer for the pointwise
convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
depthwise_constraint: Optional projection function to be applied to the
depthwise kernel after being updated by an `Optimizer` (e.g. used for
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
pointwise_constraint: Optional projection function to be applied to the
pointwise kernel after being updated by an `Optimizer`.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
Output tensor.
Raises:
ValueError: if eager execution is enabled.
"""
layer = SeparableConv2D(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
dilation_rate=dilation_rate,
depth_multiplier=depth_multiplier,
activation=activation,
use_bias=use_bias,
depthwise_initializer=depthwise_initializer,
pointwise_initializer=pointwise_initializer,
bias_initializer=bias_initializer,
depthwise_regularizer=depthwise_regularizer,
pointwise_regularizer=pointwise_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
depthwise_constraint=depthwise_constraint,
pointwise_constraint=pointwise_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
_reuse=reuse,
_scope=name)
return layer.apply(inputs)
@tf_export('layers.Conv2DTranspose')
class Conv2DTranspose(Conv2D):
"""Transposed 2D convolution layer (sometimes called 2D Deconvolution).
The need for transposed convolutions generally arises
from the desire to use a transformation going in the opposite direction
of a normal convolution, i.e., from something that has the shape of the
output of some convolution to something that has the shape of its input
while maintaining a connectivity pattern that is compatible with
said convolution.
Arguments:
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A tuple or list of 2 positive integers specifying the spatial
dimensions of the filters. Can be a single integer to specify the same
value for all spatial dimensions.
strides: A tuple or list of 2 positive integers specifying the strides
of the convolution. Can be a single integer to specify the same value for
all spatial dimensions.
padding: one of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, height, width, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, height, width)`.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self, filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(Conv2DTranspose, self).__init__(
filters,
kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
**kwargs)
self.input_spec = base.InputSpec(ndim=4)
def build(self, input_shape):
if len(input_shape) != 4:
raise ValueError('Inputs should have rank 4. Received input shape: ' +
str(input_shape))
if self.data_format == 'channels_first':
channel_axis = 1
else:
channel_axis = -1
if input_shape[channel_axis] is None:
raise ValueError('The channel dimension of the inputs '
'should be defined. Found `None`.')
input_dim = input_shape[channel_axis]
self.input_spec = base.InputSpec(ndim=4, axes={channel_axis: input_dim})
kernel_shape = self.kernel_size + (self.filters, input_dim)
self.kernel = self.add_variable(name='kernel',
shape=kernel_shape,
initializer=self.kernel_initializer,
regularizer=self.kernel_regularizer,
constraint=self.kernel_constraint,
trainable=True,
dtype=self.dtype)
if self.use_bias:
self.bias = self.add_variable(name='bias',
shape=(self.filters,),
initializer=self.bias_initializer,
regularizer=self.bias_regularizer,
constraint=self.bias_constraint,
trainable=True,
dtype=self.dtype)
else:
self.bias = None
self.built = True
def call(self, inputs):
inputs_shape = array_ops.shape(inputs)
batch_size = inputs_shape[0]
if self.data_format == 'channels_first':
c_axis, h_axis, w_axis = 1, 2, 3
else:
c_axis, h_axis, w_axis = 3, 1, 2
height, width = inputs_shape[h_axis], inputs_shape[w_axis]
kernel_h, kernel_w = self.kernel_size
stride_h, stride_w = self.strides
# Infer the dynamic output shape:
out_height = utils.deconv_output_length(height,
kernel_h,
self.padding,
stride_h)
out_width = utils.deconv_output_length(width,
kernel_w,
self.padding,
stride_w)
if self.data_format == 'channels_first':
output_shape = (batch_size, self.filters, out_height, out_width)
strides = (1, 1, stride_h, stride_w)
else:
output_shape = (batch_size, out_height, out_width, self.filters)
strides = (1, stride_h, stride_w, 1)
output_shape_tensor = array_ops.stack(output_shape)
outputs = nn.conv2d_transpose(
inputs,
self.kernel,
output_shape_tensor,
strides,
padding=self.padding.upper(),
data_format=utils.convert_data_format(self.data_format, ndim=4))
if not context.executing_eagerly():
# Infer the static output shape:
out_shape = inputs.get_shape().as_list()
out_shape[c_axis] = self.filters
out_shape[h_axis] = utils.deconv_output_length(out_shape[h_axis],
kernel_h,
self.padding,
stride_h)
out_shape[w_axis] = utils.deconv_output_length(out_shape[w_axis],
kernel_w,
self.padding,
stride_w)
outputs.set_shape(out_shape)
if self.use_bias:
outputs = nn.bias_add(
outputs,
self.bias,
data_format=utils.convert_data_format(self.data_format, ndim=4))
if self.activation is not None:
return self.activation(outputs)
return outputs
def compute_output_shape(self, input_shape):
input_shape = tensor_shape.TensorShape(input_shape).as_list()
output_shape = list(input_shape)
if self.data_format == 'channels_first':
c_axis, h_axis, w_axis = 1, 2, 3
else:
c_axis, h_axis, w_axis = 3, 1, 2
kernel_h, kernel_w = self.kernel_size
stride_h, stride_w = self.strides
output_shape[c_axis] = self.filters
output_shape[h_axis] = utils.deconv_output_length(
output_shape[h_axis], kernel_h, self.padding, stride_h)
output_shape[w_axis] = utils.deconv_output_length(
output_shape[w_axis], kernel_w, self.padding, stride_w)
return tensor_shape.TensorShape(output_shape)
@tf_export('layers.conv2d_transpose')
def conv2d_transpose(inputs,
filters,
kernel_size,
strides=(1, 1),
padding='valid',
data_format='channels_last',
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None):
"""Functional interface for transposed 2D convolution layer.
The need for transposed convolutions generally arises
from the desire to use a transformation going in the opposite direction
of a normal convolution, i.e., from something that has the shape of the
output of some convolution to something that has the shape of its input
while maintaining a connectivity pattern that is compatible with
said convolution.
Arguments:
inputs: Input tensor.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A tuple or list of 2 positive integers specifying the spatial
dimensions of the filters. Can be a single integer to specify the same
value for all spatial dimensions.
strides: A tuple or list of 2 positive integers specifying the strides
of the convolution. Can be a single integer to specify the same value for
all spatial dimensions.
padding: one of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, height, width, channels)` while `channels_first` corresponds to
inputs with shape `(batch, channels, height, width)`.
activation: Activation function. Set it to `None` to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If `None`, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
Output tensor.
Raises:
ValueError: if eager execution is enabled.
"""
layer = Conv2DTranspose(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
dtype=inputs.dtype.base_dtype,
_reuse=reuse,
_scope=name)
return layer.apply(inputs)
@tf_export('layers.Conv3DTranspose')
class Conv3DTranspose(Conv3D):
"""Transposed 3D convolution layer (sometimes called 3D Deconvolution).
Arguments:
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: An integer or tuple/list of 3 integers, specifying the
depth, height and width of the 3D convolution window.
Can be a single integer to specify the same value for all spatial
dimensions.
strides: An integer or tuple/list of 3 integers, specifying the strides
of the convolution along the depth, height and width.
Can be a single integer to specify the same value for all spatial
dimensions.
padding: One of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, depth, height, width, channels)` while `channels_first`
corresponds to inputs with shape
`(batch, channels, depth, height, width)`.
activation: Activation function. Set it to `None` to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If `None`, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
"""
def __init__(self,
filters,
kernel_size,
strides=(1, 1, 1),
padding='valid',
data_format='channels_last',
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
**kwargs):
super(Conv3DTranspose, self).__init__(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
**kwargs)
self.input_spec = base.InputSpec(ndim=5)
def build(self, input_shape):
if len(input_shape) != 5:
raise ValueError('Inputs should have rank 5, received input shape:',
str(input_shape))
if self.data_format == 'channels_first':
channel_axis = 1
else:
channel_axis = -1
if input_shape[channel_axis] is None:
raise ValueError('The channel dimension of the inputs '
'should be defined, found None: ' + str(input_shape))
input_dim = input_shape[channel_axis]
kernel_shape = self.kernel_size + (self.filters, input_dim)
self.kernel = self.add_variable(
'kernel',
shape=kernel_shape,
initializer=self.kernel_initializer,
regularizer=self.kernel_regularizer,
constraint=self.kernel_constraint,
trainable=True,
dtype=self.dtype)
if self.use_bias:
self.bias = self.add_variable(
'bias',
shape=(self.filters,),
initializer=self.bias_initializer,
regularizer=self.bias_regularizer,
constraint=self.bias_constraint,
trainable=True,
dtype=self.dtype)
else:
self.bias = None
self.built = True
def call(self, inputs):
inputs_shape = array_ops.shape(inputs)
batch_size = inputs_shape[0]
if self.data_format == 'channels_first':
c_axis, d_axis, h_axis, w_axis = 1, 2, 3, 4
else:
c_axis, d_axis, h_axis, w_axis = 4, 1, 2, 3
self.input_spec = base.InputSpec(ndim=5,
axes={c_axis: inputs_shape[c_axis]})
depth = inputs_shape[d_axis]
height = inputs_shape[h_axis]
width = inputs_shape[w_axis]
kernel_d, kernel_h, kernel_w = self.kernel_size
stride_d, stride_h, stride_w = self.strides
# Infer the dynamic output shape:
out_depth = utils.deconv_output_length(depth,
kernel_d,
self.padding,
stride_d)
out_height = utils.deconv_output_length(height,
kernel_h,
self.padding,
stride_h)
out_width = utils.deconv_output_length(width,
kernel_w,
self.padding,
stride_w)
if self.data_format == 'channels_first':
output_shape = (batch_size, self.filters, out_depth, out_height,
out_width)
strides = (1, 1, stride_d, stride_h, stride_w)
else:
output_shape = (batch_size, out_depth, out_height, out_width,
self.filters)
strides = (1, stride_d, stride_h, stride_w, 1)
output_shape_tensor = array_ops.stack(output_shape)
outputs = nn.conv3d_transpose(
inputs,
self.kernel,
output_shape_tensor,
strides,
data_format=utils.convert_data_format(self.data_format, ndim=5),
padding=self.padding.upper())
if not context.executing_eagerly():
# Infer the static output shape:
out_shape = inputs.get_shape().as_list()
out_shape[c_axis] = self.filters
out_shape[d_axis] = utils.deconv_output_length(out_shape[d_axis],
kernel_d,
self.padding,
stride_d)
out_shape[h_axis] = utils.deconv_output_length(out_shape[h_axis],
kernel_h,
self.padding,
stride_h)
out_shape[w_axis] = utils.deconv_output_length(out_shape[w_axis],
kernel_w,
self.padding,
stride_w)
outputs.set_shape(out_shape)
if self.use_bias:
outputs_shape = outputs.shape.as_list()
if outputs_shape[0] is None:
outputs_shape[0] = -1
if self.data_format == 'channels_first':
outputs_4d = array_ops.reshape(outputs, [
outputs_shape[0], outputs_shape[1],
outputs_shape[2] * outputs_shape[3], outputs_shape[4]
])
else:
outputs_4d = array_ops.reshape(outputs, [
outputs_shape[0], outputs_shape[1] * outputs_shape[2],
outputs_shape[3], outputs_shape[4]
])
outputs_4d = nn.bias_add(
outputs_4d,
self.bias,
data_format=utils.convert_data_format(self.data_format, ndim=4))
outputs = array_ops.reshape(outputs_4d, outputs_shape)
if self.activation is not None:
return self.activation(outputs)
return outputs
def compute_output_shape(self, input_shape):
input_shape = tensor_shape.TensorShape(input_shape).as_list()
output_shape = list(input_shape)
if self.data_format == 'channels_first':
c_axis, d_axis, h_axis, w_axis = 1, 2, 3, 4
else:
c_axis, d_axis, h_axis, w_axis = 4, 1, 2, 3
kernel_d, kernel_h, kernel_w = self.kernel_size
stride_d, stride_h, stride_w = self.strides
output_shape[c_axis] = self.filters
output_shape[d_axis] = utils.deconv_output_length(
output_shape[d_axis], kernel_d, self.padding, stride_d)
output_shape[h_axis] = utils.deconv_output_length(
output_shape[h_axis], kernel_h, self.padding, stride_h)
output_shape[w_axis] = utils.deconv_output_length(
output_shape[w_axis], kernel_w, self.padding, stride_w)
return tensor_shape.TensorShape(output_shape)
@tf_export('layers.conv3d_transpose')
def conv3d_transpose(inputs,
filters,
kernel_size,
strides=(1, 1, 1),
padding='valid',
data_format='channels_last',
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=init_ops.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None):
"""Functional interface for transposed 3D convolution layer.
Arguments:
inputs: Input tensor.
filters: Integer, the dimensionality of the output space (i.e. the number
of filters in the convolution).
kernel_size: A tuple or list of 3 positive integers specifying the spatial
dimensions of the filters. Can be a single integer to specify the same
value for all spatial dimensions.
strides: A tuple or list of 3 positive integers specifying the strides
of the convolution. Can be a single integer to specify the same value for
all spatial dimensions.
padding: one of `"valid"` or `"same"` (case-insensitive).
data_format: A string, one of `channels_last` (default) or `channels_first`.
The ordering of the dimensions in the inputs.
`channels_last` corresponds to inputs with shape
`(batch, depth, height, width, channels)` while `channels_first`
corresponds to inputs with shape
`(batch, channels, depth, height, width)`.
activation: Activation function. Set it to None to maintain a
linear activation.
use_bias: Boolean, whether the layer uses a bias.
kernel_initializer: An initializer for the convolution kernel.
bias_initializer: An initializer for the bias vector. If None, the default
initializer will be used.
kernel_regularizer: Optional regularizer for the convolution kernel.
bias_regularizer: Optional regularizer for the bias vector.
activity_regularizer: Optional regularizer function for the output.
kernel_constraint: Optional projection function to be applied to the
kernel after being updated by an `Optimizer` (e.g. used to implement
norm constraints or value constraints for layer weights). The function
must take as input the unprojected variable and must return the
projected variable (which must have the same shape). Constraints are
not safe to use when doing asynchronous distributed training.
bias_constraint: Optional projection function to be applied to the
bias after being updated by an `Optimizer`.
trainable: Boolean, if `True` also add variables to the graph collection
`GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
name: A string, the name of the layer.
reuse: Boolean, whether to reuse the weights of a previous layer
by the same name.
Returns:
Output tensor.
Raises:
ValueError: if eager execution is enabled.
"""
layer = Conv3DTranspose(
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=padding,
data_format=data_format,
activation=activation,
use_bias=use_bias,
kernel_initializer=kernel_initializer,
bias_initializer=bias_initializer,
kernel_regularizer=kernel_regularizer,
bias_regularizer=bias_regularizer,
activity_regularizer=activity_regularizer,
kernel_constraint=kernel_constraint,
bias_constraint=bias_constraint,
trainable=trainable,
name=name,
dtype=inputs.dtype.base_dtype,
_reuse=reuse,
_scope=name)
return layer.apply(inputs)
# Aliases
Convolution1D = Conv1D
Convolution2D = Conv2D
Convolution3D = Conv3D
SeparableConvolution2D = SeparableConv2D
Convolution2DTranspose = Deconvolution2D = Deconv2D = Conv2DTranspose
Convolution3DTranspose = Deconvolution3D = Deconv3D = Conv3DTranspose
convolution1d = conv1d
convolution2d = conv2d
convolution3d = conv3d
separable_convolution2d = separable_conv2d
convolution2d_transpose = deconvolution2d = deconv2d = conv2d_transpose
convolution3d_transpose = deconvolution3d = deconv3d = conv3d_transpose