def __init__(self,
filters,
data_format=None,
activation=tf.nn.relu,
*args,
**kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {
'strides': (2, 2, 2),
'padding': 'same',
'data_format': data_format,
'filters': filters
}
layers = [
Conv3D(kernel_size=(9, 9, 9),
use_bias=True,
activation=activation,
**params),
Conv3D(kernel_size=(5, 5, 5),
use_bias=True,
activation=activation,
**params),
Conv3D(kernel_size=(5, 5, 5),
use_bias=False,
activation=None,
**params)
]
super(AnalysisTransformV1, self).__init__(layers, *args, **kwargs)
def __init__(self,
filters,
data_format=None,
kernel_size=(3, 3, 3),
strides=(2, 2, 2),
activation=tf.nn.relu,
*args,
**kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {
'padding': 'same',
'data_format': data_format,
'use_bias': True,
'activation': activation,
'filters': filters,
'kernel_size': kernel_size
}
layers = [
Conv3DTranspose(strides=strides, **params),
Conv3DTranspose(**params),
Conv3DTranspose(**params)
]
super(SynthesisBlock, self).__init__(layers,
*args,
data_format=data_format,
**kwargs)
def __init__(self,
filters,
data_format=None,
kernel_size=(3, 3, 3),
activation=tf.nn.relu,
residual_mode='add',
*args,
**kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {
'kernel_size': kernel_size,
'activation': activation,
'data_format': data_format,
'residual_mode': residual_mode
}
layers = [
SynthesisBlock(filters, **params),
SynthesisBlock(filters // 2, **params),
SynthesisBlock(filters // 4, **params),
Conv3DTranspose(1,
kernel_size,
padding="same",
use_bias=True,
activation=activation,
data_format=data_format)
]
super(SynthesisTransformProgressiveV2,
self).__init__(layers, *args, **kwargs)
def __init__(self,
filters,
data_format=None,
kernel_size=(3, 3, 3),
activation=tf.nn.relu,
residual_mode='add',
*args,
**kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {
'kernel_size': kernel_size,
'activation': activation,
'data_format': data_format,
'residual_mode': residual_mode
}
layers = [
AnalysisBlock(filters // 2, **params),
AnalysisBlock(filters, **params),
AnalysisBlock(filters, **params),
Conv3D(filters,
kernel_size,
padding="same",
use_bias=False,
activation=None,
data_format=data_format)
]
super(AnalysisTransformV2, self).__init__(layers, *args, **kwargs)
def __init__(self, filters, data_format=None, kernel_size=(3, 3, 3), activation=tf.nn.relu, *args, **kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {'padding': 'same', 'data_format': data_format, 'filters': filters, 'kernel_size': kernel_size}
layers = [Conv3D(use_bias=True, activation=activation, **params),
Conv3D(use_bias=True, activation=activation, strides=(2, 2, 2), **params),
Conv3D(use_bias=False, activation=None, **params)]
super(HyperAnalysisTransform, self).__init__(layers, *args, **kwargs)
def __init__(self, filters, data_format=None, kernel_size=(3, 3, 3), strides=(2, 2, 2), activation=tf.nn.relu, *args, **kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {'padding': 'same', 'data_format': data_format,
'filters': filters, 'kernel_size': kernel_size, 'use_bias': True}
layers = [Conv3D(strides=strides, activation=activation() **params),
AnalysisBlockV3Base(filters=filters, data_format=data_format),
AnalysisBlockV3Base(filters=filters, data_format=data_format)]
super(AnalysisBlockV3, self).__init__(layers, *args, data_format=data_format, **kwargs)
def __init__(self, filters, data_format=None, activation=tf.nn.relu, *args, **kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {'strides': (2, 2, 2), 'padding': 'same', 'data_format': data_format, 'use_bias': True,
'activation': activation}
layers = [Conv3DTranspose(filters, (5, 5, 5), **params),
Conv3DTranspose(filters, (5, 5, 5), **params),
Conv3DTranspose(1, (9, 9, 9), **params)]
super(SynthesisTransformV1, self).__init__(layers, *args, **kwargs)
def __init__(self, filters, data_format=None, activation=ACTIVATION, *args, **kwargs):
data_format = conv_utils.normalize_data_format(data_format)
params = {'activation': activation, 'data_format': data_format}
layers = [AnalysisBlock(filters // 4, **params),
AnalysisBlockV4(filters // 2, **params),
AnalysisBlockV4(filters, **params),
Conv3D(filters, (3, 3, 3), padding="same", use_bias=False, activation=None,
data_format=data_format)]
super(AnalysisTransformProgressiveV6, self).__init__(layers, *args, **kwargs)
def __init__(self, filters, data_format=None, activation=ACTIVATION, *args, **kwargs):
self.data_format = conv_utils.normalize_data_format(data_format)
params = {'padding': 'same', 'data_format': data_format, 'use_bias': True}
self.paths = [[Conv3D(kernel_size=(1,1,1), activation=activation(), filters=filters // 4, **params)],
[Conv3D(kernel_size=(1,1,1), activation=activation(), filters=filters // 2, **params),
[Conv3D(kernel_size=(1,1,3), activation=activation(), filters=filters // 4, **params), Conv3D(kernel_size=(3,3,1), activation=activation(), filters=filters // 4, **params),
Conv3D(kernel_size=(3,1,1), activation=activation(), filters=filters // 4, **params), Conv3D(kernel_size=(1,3,3), activation=activation(), filters=filters // 4, **params),
Conv3D(kernel_size=(1,3,1), activation=activation(), filters=filters // 4, **params), Conv3D(kernel_size=(3,1,3), activation=activation(), filters=filters // 4, **params)]]]
super(AnalysisBlockV3Base, self).__init__(Self.paths, *args, **kwargs)
def __init__(self,
layers,
residual_mode='add',
data_format=None,
*args,
**kwargs):
super(ResidualLayer, self).__init__(*args, **kwargs)
assert residual_mode in ('add', 'concat')
self._layers = layers
self.residual_mode = residual_mode
self.data_format = conv_utils.normalize_data_format(data_format)
def __init__(self,
pool_size,
strides,
padding='valid',
data_format=None,
name=None,
**kwargs):
super(MaxPoolWithArgmax2D, self).__init__(name=name, **kwargs)
if data_format is None:
data_format = tf.keras.backend.image_data_format()
if strides is None:
strides = pool_size
self.pool_size = conv_utils.normalize_tuple(pool_size, 2, 'pool_size')
self.strides = conv_utils.normalize_tuple(strides, 2, 'strides')
self.padding = conv_utils.normalize_padding(padding)
self.data_format = conv_utils.normalize_data_format(data_format)
self.input_spec = tf.keras.layers.InputSpec(ndim=4)
def __init__(self, data_format='channels_last', name=None, **kwargs):
super(MaxUnpool2D, self).__init__(**kwargs)
if data_format is None:
data_format = tf.keras.backend.image_data_format()
self.data_format = conv_utils.normalize_data_format(data_format)
self.input_spec = tf.keras.layers.InputSpec(min_ndim=2, max_ndim=4)
