def __init__(self, input_tensor, is_training, reuse):
# net = layers.Input(shape=input_tensor)
net = input_tensor
with tf.variable_scope('Encoder'):
with tf.variable_scope('Convolution'):
# Layer 1: A conventional Conv2D layer
net = layers.Conv2D(filters=16,
kernel_size=5,
strides=1,
padding='same',
activation='relu',
name='conv1')(net)
self.conv1 = net
# Reshape layer to be 1 capsule x [filters] atoms
_, H, W, C = net.get_shape()
net = layers.Reshape((H.value, W.value, 1, C.value))(net)
# Layer 1: Primary Capsule: Conv cap with routing 1
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=2,
num_atoms=8,
strides=2,
padding='same',
routings=1,
name='primarycaps')(net)
self.primary_caps = net
# Layer 2: Convolutional Capsules
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=4,
num_atoms=8,
strides=2,
padding='same',
routings=3,
name='conv_cap_2')(net)
self.conv_cap_2 = net
# Layer 3: Convolutional Capsules
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=8,
num_atoms=16,
strides=2,
padding='same',
routings=3,
name='conv_cap_3')(net)
self.conv_cap_3 = net
# Layer 4: Convolutional Capsules
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=8,
num_atoms=32,
strides=2,
padding='same',
routings=3,
name='conv_cap_4')(net)
self.output = net
def __init__(self, input_tensor, is_training, reuse):
self.input_tensor = input_tensor
with tf.variable_scope('Encoder'):
# conv1 = layers.Conv2D(filters=16, kernel_size=5, strides=1, padding='same', activation='relu', name='conv1')(x)
_, H, W, C = self.input_tensor.get_shape()
net = layers.Reshape(
(H.value, W.value, 1, C.value))(self.input_tensor)
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=2,
num_atoms=2,
strides=2,
padding='same',
routings=3,
name='caps_conv1')(net)
net = mf.batch_norm(net, is_training=is_training, reuse=reuse)
self.caps_conv1 = net
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=4,
num_atoms=4,
strides=2,
padding='same',
routings=3,
name='caps_conv2')(net)
net = mf.batch_norm(net, is_training=is_training, reuse=reuse)
self.caps_conv2 = net
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=8,
num_atoms=8,
strides=2,
padding='same',
routings=3,
name='caps_conv3')(net)
net = mf.batch_norm(net, is_training=is_training, reuse=reuse)
self.caps_conv3 = net
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=16,
num_atoms=16,
strides=2,
padding='same',
routings=3,
name='caps_conv4')(net)
#self.caps_conv4 = net
#net = capsule_layers.ConvCapsuleLayer(kernel_size=5, num_capsule=32, num_atoms=2, strides=2, padding='same',
# routings=3, name='caps_conv5')(net)
#self.caps_conv5 = net
#net = capsule_layers.ConvCapsuleLayer(kernel_size=5, num_capsule=64, num_atoms=2, strides=2, padding='same',
# routings=3, name='caps_conv6')(net)
self.output = net
def __init__(self, mixed_mag, name='SegCaps_CapsNetBasic'):
"""
input_tensor: Tensor with shape [batch_size, height, width, channels]
is_training: Boolean - should the model be trained on the current input or not
name: Model instance name
"""
with tf.variable_scope(name):
self.mixed_mag = mixed_mag
with tf.variable_scope('Convolution'):
conv1 = mf.conv(mixed_mag,
filters=128,
kernel_size=5,
stride=(1, 1))
# Reshape layer to be 1 capsule x [filters] atoms
_, H, W, C = conv1.get_shape()
conv1 = layers.Reshape((H.value, W.value, 1, C.value))(conv1)
self.conv1 = conv1
with tf.variable_scope('Primary_Caps'):
primary_caps = capsule_layers.ConvCapsuleLayer(
kernel_size=5,
num_capsule=8,
num_atoms=32,
strides=1,
padding='same',
routings=1,
name='primarycaps')(conv1)
self.primary_caps = primary_caps
with tf.variable_scope('Seg_Caps'):
seg_caps = capsule_layers.ConvCapsuleLayer(
kernel_size=1,
num_capsule=1,
num_atoms=16,
strides=1,
padding='same',
routings=3,
name='seg_caps')(primary_caps)
self.seg_caps = seg_caps
with tf.variable_scope('Reconstruction'):
reconstruction = capsule_layers.ConvCapsuleLayer(
kernel_size=1,
num_capsule=1,
num_atoms=1,
strides=1,
padding='same',
routings=3,
name='seg_caps')(primary_caps)
reconstruction = tf.squeeze(reconstruction, -1)
self.output = reconstruction
def __init__(self, mixed_mag, name):
"""
A basic capsule network operating on magnitude spectrograms.
"""
with tf.variable_scope(name):
self.mixed_mag = mixed_mag
with tf.variable_scope('Convolution'):
net = mf.conv(mixed_mag,
filters=128,
kernel_size=5,
stride=(1, 1))
# Reshape layer to be 1 capsule x [filters] atoms
_, H, W, C = net.get_shape()
net = layers.Reshape((H.value, W.value, 1, C.value))(net)
self.conv1 = net
with tf.variable_scope('Primary_Caps'):
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=8,
num_atoms=8,
strides=1,
padding='same',
routings=1,
name='primarycaps')(net)
self.primary_caps = net
with tf.variable_scope('Seg_Caps'):
net = capsule_layers.ConvCapsuleLayer(kernel_size=1,
num_capsule=1,
num_atoms=8,
strides=1,
padding='same',
routings=3,
name='seg_caps')(net)
self.seg_caps = net
with tf.variable_scope('Reconstruction'):
net = capsule_layers.ConvCapsuleLayer(
kernel_size=1,
num_capsule=1,
num_atoms=1,
strides=1,
padding='same',
routings=3,
name='reconstruction')(net)
net = tf.squeeze(net, -1)
self.output = net
def __init__(self, input_tensor, name):
"""
A basic capsule network operating on magnitude spectrograms.
"""
with tf.variable_scope(name):
self.input_tensor = input_tensor
if tf.rank(self.input_tensor) == 3:
self.out_depth = 1
else:
self.out_depth = input_tensor.shape[3].value
with tf.variable_scope('layer_1'):
net = mf.conv(input_tensor,
filters=128,
kernel_size=5,
stride=(1, 1))
# Reshape layer to be 1 capsule x [filters] atoms
_, H, W, C = net.get_shape()
net = layers.Reshape((H.value, W.value, 1, C.value))(net)
self.conv1 = net
net = capsule_layers.ConvCapsuleLayer(kernel_size=5,
num_capsule=8,
num_atoms=16,
strides=1,
padding='same',
routings=1,
name='layer_2')(net)
self.primary_caps = net
net = capsule_layers.ConvCapsuleLayer(kernel_size=1,
num_capsule=1,
num_atoms=16,
strides=1,
padding='same',
routings=3,
name='layer_3')(net)
self.seg_caps = net
net = capsule_layers.ConvCapsuleLayer(kernel_size=1,
num_capsule=self.out_depth,
num_atoms=1,
strides=1,
padding='same',
routings=3,
name='mask')(net)
net = tf.squeeze(net, -1)
self.output = net
def __init__(self, input_tensor):
net = input_tensor
with tf.variable_scope('encoder'):
with tf.variable_scope('layer-1'):
# Reshape layer to be 1 capsule x [filters] atoms
_, self.H, self.W, self.C = net.get_shape()
net = layers.Reshape((self.H.value, self.W.value, 1, self.C.value))(net)
self.l1 = net
with tf.variable_scope('layer-2'):
net = capsule_layers.ConvCapsuleLayer(kernel_size=5, num_capsule=8, num_atoms=2, strides=2,
padding='same',
routings=1, name='primarycaps')(net)
self.l2 = net
with tf.variable_scope('layer-3'):
net = capsule_layers.ConvCapsuleLayer(kernel_size=5, num_capsule=32, num_atoms=2, strides=2,
padding='same',
routings=1, name='primarycaps')(net)
self.output = net
def __init__(self, input_tensor, encoder, is_training, reuse):
net = input_tensor
with tf.variable_scope('Decoder'):
with tf.variable_scope('UpCaps_1'):
# Layer 1 Up: Deconvolutional capsules, skip connection, convolutional capsules
net = capsule_layers.DeconvCapsuleLayer(
kernel_size=4,
num_capsule=8,
num_atoms=32,
upsamp_type='deconv',
scaling=2,
padding='same',
routings=3,
name='deconv_cap_1_1')(net)
self.upcap_1_1 = net
net = layers.Concatenate(
axis=-2, name='up_1')([net, encoder.conv_cap_3_1])
net = capsule_layers.ConvCapsuleLayer(
kernel_size=5,
num_capsule=4,
num_atoms=32,
strides=1,
padding='same',
routings=3,
name='deconv_cap_1_2')(net)
self.upcap_1_2 = net
with tf.variable_scope('UpCaps_2'):
# Layer 2 Up: Deconvolutional capsules, skip connection, convolutional capsules
net = capsule_layers.DeconvCapsuleLayer(
kernel_size=4,
num_capsule=4,
num_atoms=16,
upsamp_type='deconv',
scaling=2,
padding='same',
routings=3,
name='deconv_cap_2_1')(net)
self.upcap_2_1 = net
net = layers.Concatenate(
axis=-2, name='up_2')([net, encoder.conv_cap_2_1])
net = capsule_layers.ConvCapsuleLayer(
kernel_size=5,
num_capsule=4,
num_atoms=16,
strides=1,
padding='same',
routings=3,
name='deconv_cap_2_2')(net)
self.upcap_2_2 = net
with tf.variable_scope('UpCaps_3'):
# Layer 3 Up: Deconvolutional capsules, skip connection
net = capsule_layers.DeconvCapsuleLayer(
kernel_size=4,
num_capsule=2,
num_atoms=16,
upsamp_type='deconv',
scaling=2,
padding='same',
routings=3,
name='deconv_cap_3_1')(net)
self.upcap_3_1 = net
net = layers.Concatenate(axis=-2,
name='up_3')([net, encoder.conv1])
with tf.variable_scope('Reconstruction'):
# Layer 4: Reconstruction - Convolutional Capsule: 1x1, 3x conventional Conv2D layers
net = capsule_layers.ConvCapsuleLayer(kernel_size=1,
num_capsule=1,
num_atoms=16,
strides=1,
padding='same',
routings=3,
name='seg_caps')(net)
_, H, W, C, D = net.get_shape()
net = layers.Reshape((H.value, W.value, D.value))(net)
net = layers.Conv2D(filters=64,
kernel_size=1,
padding='same',
kernel_initializer='he_normal',
activation='relu',
name='recon_1')(net)
net = layers.Conv2D(filters=128,
kernel_size=1,
padding='same',
kernel_initializer='he_normal',
activation='relu',
name='recon_2')(net)
net = layers.Conv2D(filters=1,
kernel_size=1,
padding='same',
kernel_initializer='he_normal',
activation='sigmoid',
name='out_recon')(net)
self.output = net
