def C2D_fcn_16_body(x, training, scope_name='classifier', scope_reuse=False):
with tf.variable_scope(scope_name) as scope:
if scope_reuse:
scope.reuse_variables()
conv1_1 = layers.conv2D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool2D_layer(conv1_1)
conv2_1 = layers.conv2D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool2D_layer(conv2_1)
conv3_1 = layers.conv2D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv2D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool2D_layer(conv3_2)
conv4_1 = layers.conv2D_layer(pool3, 'conv4_1', num_filters=128)
conv4_2 = layers.conv2D_layer(conv4_1, 'conv4_2', num_filters=128)
pool4 = layers.maxpool2D_layer(conv4_2)
conv5_1 = layers.conv2D_layer(pool4, 'conv5_1', num_filters=256)
conv5_2 = layers.conv2D_layer(conv5_1, 'conv5_2', num_filters=256)
convD_1 = layers.conv2D_layer(conv5_2, 'convD_1', num_filters=256)
logits = tf.reduce_mean(convD_1, axis=(1, 2))
return logits
def C3D_fcn_16_2D_conditioned_with_delta(x,
training,
scope_name='critic',
scope_reuse=False):
with tf.variable_scope(scope_name) as scope:
if scope_reuse:
scope.reuse_variables()
# discrimnator is conditioned on the difference map
condition = x[:, :, :, 2:3]
delta_img = x[:, :, :1:2]
delta = delta_img[:, 0, 0, :]
x = x[:, :, :, 0:1]
conv1_1 = layers.conv2D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool2D_layer(conv1_1)
conv2_1 = layers.conv2D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool2D_layer(conv2_1)
conv3_1 = layers.conv2D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv2D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool2D_layer(conv3_2)
conv4_1 = layers.conv2D_layer(pool3, 'conv4_1', num_filters=128)
conv4_2 = layers.conv2D_layer(conv4_1, 'conv4_2', num_filters=128)
pool4 = layers.maxpool2D_layer(conv4_2)
conv5_1 = layers.conv2D_layer(pool4, 'conv5_1', num_filters=256)
conv5_2 = layers.conv2D_layer(conv5_1, 'conv5_2', num_filters=256)
convD_1 = layers.conv2D_layer(conv5_2, 'convD_1', num_filters=256)
convD_2 = layers.conv2D_layer(convD_1,
'convD_2',
num_filters=1,
kernel_size=(1, 1, 1),
activation=tf.identity)
logits = layers.averagepool2D_layer(convD_2, name='diagnosis_avg')
condition_logits = C3D_fcn_16_2D(condition,
training,
scope_name="critic_condition",
scope_reuse=scope_reuse)
all_logits = tf.concat([logits, condition_logits, delta], axis=-1)
decision = tf.layers.dense(all_logits, 1, reuse=scope_reuse)
return decision
def C3D_fcn_16_2D_conditioned(x,
training,
scope_name='critic',
scope_reuse=False):
with tf.variable_scope(scope_name) as scope:
if scope_reuse:
scope.reuse_variables()
# check if last channel larger than 1
condition = x[:, :, :, 1:2]
x = x[:, :, :, 0:1]
conv1_1 = layers.conv2D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool2D_layer(conv1_1)
conv2_1 = layers.conv2D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool2D_layer(conv2_1)
conv3_1 = layers.conv2D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv2D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool2D_layer(conv3_2)
conv4_1 = layers.conv2D_layer(pool3, 'conv4_1', num_filters=128)
conv4_2 = layers.conv2D_layer(conv4_1, 'conv4_2', num_filters=128)
pool4 = layers.maxpool2D_layer(conv4_2)
conv5_1 = layers.conv2D_layer(pool4, 'conv5_1', num_filters=256)
conv5_2 = layers.conv2D_layer(conv5_1, 'conv5_2', num_filters=256)
convD_1 = layers.conv2D_layer(conv5_2, 'convD_1', num_filters=256)
convD_2 = layers.conv2D_layer(convD_1,
'convD_2',
num_filters=1,
kernel_size=(1, 1, 1),
activation=tf.identity)
logits = layers.averagepool2D_layer(convD_2, name='diagnosis_avg')
condition_logits = C3D_fcn_16_2D(condition,
training,
scope_name="critic_condition",
scope_reuse=scope_reuse)
all_logits = tf.concat([logits, condition_logits], axis=-1)
decision = tf.layers.dense(all_logits, 1, reuse=scope_reuse)
return decision
def C3D_fcn_16_2D_body(x, training, scope_name='critic', scope_reuse=False):
with tf.variable_scope(scope_name) as scope:
if scope_reuse:
scope.reuse_variables()
conv1_1 = layers.conv2D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool2D_layer(conv1_1)
conv2_1 = layers.conv2D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool2D_layer(conv2_1)
conv3_1 = layers.conv2D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv2D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool2D_layer(conv3_2)
return pool3
def C3D_fcn_16_2D(x, training, scope_name='critic', scope_reuse=False):
with tf.variable_scope(scope_name) as scope:
if scope_reuse:
scope.reuse_variables()
conv1_1 = layers.conv2D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool2D_layer(conv1_1)
conv2_1 = layers.conv2D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool2D_layer(conv2_1)
conv3_1 = layers.conv2D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv2D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool2D_layer(conv3_2)
conv4_1 = layers.conv2D_layer(pool3, 'conv4_1', num_filters=128)
conv4_2 = layers.conv2D_layer(conv4_1, 'conv4_2', num_filters=128)
pool4 = layers.maxpool2D_layer(conv4_2)
conv5_1 = layers.conv2D_layer(pool4, 'conv5_1', num_filters=256)
conv5_2 = layers.conv2D_layer(conv5_1, 'conv5_2', num_filters=256)
convD_1 = layers.conv2D_layer(conv5_2, 'convD_1', num_filters=256)
convD_2 = layers.conv2D_layer(convD_1,
'convD_2',
num_filters=1,
kernel_size=(1, 1, 1),
activation=tf.identity)
logits = layers.averagepool2D_layer(convD_2, name='diagnosis_avg')
return logits
def unet_16_2D_bn(x, training, scope_name='generator'):
n_ch_0 = 16
with tf.variable_scope(scope_name):
conv1_1 = layers.conv2D_layer_bn(x,
'conv1_1',
num_filters=n_ch_0,
training=training)
conv1_2 = layers.conv2D_layer_bn(conv1_1,
'conv1_2',
num_filters=n_ch_0,
training=training)
pool1 = layers.maxpool2D_layer(conv1_2)
conv2_1 = layers.conv2D_layer_bn(pool1,
'conv2_1',
num_filters=n_ch_0 * 2,
training=training)
conv2_2 = layers.conv2D_layer_bn(conv2_1,
'conv2_2',
num_filters=n_ch_0 * 2,
training=training)
pool2 = layers.maxpool2D_layer(conv2_2)
conv3_1 = layers.conv2D_layer_bn(pool2,
'conv3_1',
num_filters=n_ch_0 * 4,
training=training)
conv3_2 = layers.conv2D_layer_bn(conv3_1,
'conv3_2',
num_filters=n_ch_0 * 4,
training=training)
pool3 = layers.maxpool2D_layer(conv3_2)
conv4_1 = layers.conv2D_layer_bn(pool3,
'conv4_1',
num_filters=n_ch_0 * 8,
training=training)
conv4_2 = layers.conv2D_layer_bn(conv4_1,
'conv4_2',
num_filters=n_ch_0 * 8,
training=training)
upconv3 = layers.deconv2D_layer_bn(conv4_2,
name='upconv3',
num_filters=n_ch_0,
training=training)
concat3 = layers.crop_and_concat_layer_fixed([upconv3, conv3_2],
axis=-1)
conv5_1 = layers.conv2D_layer_bn(concat3,
'conv5_1',
num_filters=n_ch_0 * 4,
training=training)
conv5_2 = layers.conv2D_layer_bn(conv5_1,
'conv5_2',
num_filters=n_ch_0 * 4,
training=training)
upconv2 = layers.deconv2D_layer_bn(conv5_2,
name='upconv2',
num_filters=n_ch_0,
training=training)
concat2 = layers.crop_and_concat_layer_fixed([upconv2, conv2_2],
axis=-1)
conv6_1 = layers.conv2D_layer_bn(concat2,
'conv6_1',
num_filters=n_ch_0 * 2,
training=training)
conv6_2 = layers.conv2D_layer_bn(conv6_1,
'conv6_2',
num_filters=n_ch_0 * 2,
training=training)
upconv1 = layers.deconv2D_layer_bn(conv6_2,
name='upconv1',
num_filters=n_ch_0,
training=training)
concat1 = layers.crop_and_concat_layer_fixed([upconv1, conv1_2],
axis=-1)
conv8_1 = layers.conv2D_layer_bn(concat1,
'conv8_1',
num_filters=n_ch_0,
training=training)
conv8_2 = layers.conv2D_layer(conv8_1,
'conv8_2',
num_filters=1,
activation=tf.identity)
return conv8_2
def unet_16_2D_allow_reuse(x,
training,
scope_name='generator',
scope_reuse=True):
n_ch_0 = 16
with tf.variable_scope(scope_name, reuse=tf.AUTO_REUSE) as scope:
# if scope_reuse:
# scope.reuse_variables()
conv1_1 = layers.conv2D_layer(x, 'conv1_1', num_filters=n_ch_0)
conv1_2 = layers.conv2D_layer(conv1_1, 'conv1_2', num_filters=n_ch_0)
pool1 = layers.maxpool2D_layer(conv1_2)
conv2_1 = layers.conv2D_layer(pool1, 'conv2_1', num_filters=n_ch_0 * 2)
conv2_2 = layers.conv2D_layer(conv2_1,
'conv2_2',
num_filters=n_ch_0 * 2)
pool2 = layers.maxpool2D_layer(conv2_2)
conv3_1 = layers.conv2D_layer(pool2, 'conv3_1', num_filters=n_ch_0 * 4)
conv3_2 = layers.conv2D_layer(conv3_1,
'conv3_2',
num_filters=n_ch_0 * 4)
pool3 = layers.maxpool2D_layer(conv3_2)
conv4_1 = layers.conv2D_layer(pool3, 'conv4_1', num_filters=n_ch_0 * 8)
conv4_2 = layers.conv2D_layer(conv4_1,
'conv4_2',
num_filters=n_ch_0 * 8)
upconv3 = layers.deconv2D_layer(conv4_2,
name='upconv3',
num_filters=n_ch_0)
concat3 = layers.crop_and_concat_layer_fixed([upconv3, conv3_2],
axis=-1)
conv5_1 = layers.conv2D_layer(concat3,
'conv5_1',
num_filters=n_ch_0 * 4)
conv5_2 = layers.conv2D_layer(conv5_1,
'conv5_2',
num_filters=n_ch_0 * 4)
upconv2 = layers.deconv2D_layer(conv5_2,
name='upconv2',
num_filters=n_ch_0)
concat2 = layers.crop_and_concat_layer_fixed([upconv2, conv2_2],
axis=-1)
conv6_1 = layers.conv2D_layer(concat2,
'conv6_1',
num_filters=n_ch_0 * 2)
conv6_2 = layers.conv2D_layer(conv6_1,
'conv6_2',
num_filters=n_ch_0 * 2)
upconv1 = layers.deconv2D_layer(conv6_2,
name='upconv1',
num_filters=n_ch_0)
concat1 = layers.crop_and_concat_layer_fixed([upconv1, conv1_2],
axis=-1)
conv8_1 = layers.conv2D_layer(concat1, 'conv8_1', num_filters=n_ch_0)
conv8_2 = layers.conv2D_layer(conv8_1,
'conv8_2',
num_filters=1,
activation=tf.identity)
return conv8_2