def C3D_fcn_16_conditioned(x,
training,
scope_name='critic',
scope_reuse=False):
with tf.variable_scope(scope_name) as scope:
if scope_reuse:
scope.reuse_variables()
condition = x[:, :, :, :, 1:2]
x = x[:, :, :, :, 0:1]
conv1_1 = layers.conv3D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool3D_layer(conv1_1)
conv2_1 = layers.conv3D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool3D_layer(conv2_1)
conv3_1 = layers.conv3D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv3D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool3D_layer(conv3_2)
conv4_1 = layers.conv3D_layer(pool3, 'conv4_1', num_filters=128)
conv4_2 = layers.conv3D_layer(conv4_1, 'conv4_2', num_filters=128)
pool4 = layers.maxpool3D_layer(conv4_2)
conv5_1 = layers.conv3D_layer(pool4, 'conv5_1', num_filters=256)
conv5_2 = layers.conv3D_layer(conv5_1, 'conv5_2', num_filters=256)
convD_1 = layers.conv3D_layer(conv5_2, 'convD_1', num_filters=256)
convD_2 = layers.conv3D_layer(convD_1,
'convD_2',
num_filters=1,
kernel_size=(1, 1, 1),
activation=tf.identity)
logits = layers.averagepool3D_layer(convD_2, name='diagnosis_avg')
condition_logits = C3D_fcn_16(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_body(x,
training,
scope_name='classifier',
scope_reuse=False,
**kwargs):
with tf.variable_scope(scope_name) as scope:
if scope_reuse:
scope.reuse_variables()
conv1_1 = layers.conv3D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool3D_layer(conv1_1)
conv2_1 = layers.conv3D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool3D_layer(conv2_1)
conv3_1 = layers.conv3D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv3D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool3D_layer(conv3_2)
conv4_1 = layers.conv3D_layer(pool3, 'conv4_1', num_filters=128)
conv4_2 = layers.conv3D_layer(conv4_1, 'conv4_2', num_filters=128)
pool4 = layers.maxpool3D_layer(conv4_2)
conv5_1 = layers.conv3D_layer(pool4, 'conv5_1', num_filters=256)
conv5_2 = layers.conv3D_layer(conv5_1, 'conv5_2', num_filters=256)
n_channels = 256
if "params" in kwargs:
if "hidden_dim" in kwargs["params"]:
n_channels = kwargs["params"]["hidden_dim"]
convD_1 = layers.conv3D_layer(conv5_2,
'convD_1',
num_filters=n_channels)
logits = tf.reduce_mean(convD_1, axis=(1, 2, 3))
return logits
def C3D_fcn_16(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.conv3D_layer(x, 'conv1_1', num_filters=16)
pool1 = layers.maxpool3D_layer(conv1_1)
conv2_1 = layers.conv3D_layer(pool1, 'conv2_1', num_filters=32)
pool2 = layers.maxpool3D_layer(conv2_1)
conv3_1 = layers.conv3D_layer(pool2, 'conv3_1', num_filters=64)
conv3_2 = layers.conv3D_layer(conv3_1, 'conv3_2', num_filters=64)
pool3 = layers.maxpool3D_layer(conv3_2)
conv4_1 = layers.conv3D_layer(pool3, 'conv4_1', num_filters=128)
conv4_2 = layers.conv3D_layer(conv4_1, 'conv4_2', num_filters=128)
pool4 = layers.maxpool3D_layer(conv4_2)
conv5_1 = layers.conv3D_layer(pool4, 'conv5_1', num_filters=256)
conv5_2 = layers.conv3D_layer(conv5_1, 'conv5_2', num_filters=256)
convD_1 = layers.conv3D_layer(conv5_2, 'convD_1', num_filters=256)
convD_2 = layers.conv3D_layer(convD_1,
'convD_2',
num_filters=1,
kernel_size=(1, 1, 1),
activation=tf.identity)
logits = layers.averagepool3D_layer(convD_2, name='diagnosis_avg')
return logits
def construct_graph(self):
features = self.features
params = self.params
x = features[self.input_key]
n_filters = params["n_filters"]
filter_sizes = params["filter_sizes"]
if params["normalize_images"]:
x = self.normalize_voxels(x)
input_shape = params["input_shape"]
# Reshape to have one explicit channel
x = tf.reshape(x,
[-1, input_shape[0], input_shape[1], input_shape[2], 1])
self.x = x
current_input = x
n_ch_0 = 8
self.n_ch_0 = n_ch_0
conv1_1 = layers.conv3D_layer(x, 'conv1_1', num_filters=n_ch_0)
# conv1_2 = layers.conv3D_layer(conv1_1, 'conv1_2', num_filters=n_ch_0)
pool1 = layers.maxpool3D_layer(conv1_1)
conv2_1 = layers.conv3D_layer(pool1, 'conv2_1', num_filters=n_ch_0 * 2)
# conv2_2 = layers.conv3D_layer(conv2_1, 'conv2_2', num_filters=n_ch_0*2)
pool2 = layers.maxpool3D_layer(conv2_1)
conv3_1 = layers.conv3D_layer(pool2, 'conv3_1', num_filters=n_ch_0 * 4)
# conv3_2 = layers.conv3D_layer(conv3_1, 'conv3_2', num_filters=n_ch_0*4)
# pool3 = layers.maxpool3D_layer(conv3_2)
# conv4_1 = layers.conv3D_layer(pool3, 'conv4_1', num_filters=n_ch_0*8)
# conv4_2 = layers.conv3D_layer(conv4_1, 'conv4_2', num_filters=n_ch_0*8)
self.z = tf.contrib.layers.flatten(conv3_1)
current_input = self.z
dim_list = current_input.get_shape().as_list()[1:]
cur_dim = reduce(lambda x, y: x * y, dim_list)
self.linear_trafo = False
if cur_dim > self.get_encoding_dim():
print("Non conv layer needed")
self.linear_trafo = True
self.dim_before_linear_trafo = cur_dim
self.dim_list = dim_list
current_input = tf.contrib.layers.flatten(current_input)
W = tf.get_variable(
"non_conv_w",
shape=[cur_dim, self.get_encoding_dim()],
initializer=tf.contrib.layers.xavier_initializer(seed=40))
b = tf.get_variable("non_conv_b",
shape=[1, self.get_encoding_dim()],
initializer=tf.initializers.zeros)
self.linear_w = W
current_input = tf.add(tf.nn.relu(tf.matmul(current_input, W)), b)
self.z = current_input
self.conv1_1 = conv1_1
# self.conv1_2 = conv1_2
self.pool1 = pool1
self.conv2_1 = conv2_1
# self.conv2_2 = conv2_2
self.pool2 = pool2
self.conv3_1 = conv3_1
def unet_16_bn(x, training, scope_name='generator'):
n_ch_0 = 16
with tf.variable_scope(scope_name):
conv1_1 = layers.conv3D_layer_bn(x,
'conv1_1',
num_filters=n_ch_0,
training=training)
conv1_2 = layers.conv3D_layer_bn(conv1_1,
'conv1_2',
num_filters=n_ch_0,
training=training)
pool1 = layers.maxpool3D_layer(conv1_2)
conv2_1 = layers.conv3D_layer_bn(pool1,
'conv2_1',
num_filters=n_ch_0 * 2,
training=training)
conv2_2 = layers.conv3D_layer_bn(conv2_1,
'conv2_2',
num_filters=n_ch_0 * 2,
training=training)
pool2 = layers.maxpool3D_layer(conv2_2)
conv3_1 = layers.conv3D_layer_bn(pool2,
'conv3_1',
num_filters=n_ch_0 * 4,
training=training)
conv3_2 = layers.conv3D_layer_bn(conv3_1,
'conv3_2',
num_filters=n_ch_0 * 4,
training=training)
pool3 = layers.maxpool3D_layer(conv3_2)
conv4_1 = layers.conv3D_layer_bn(pool3,
'conv4_1',
num_filters=n_ch_0 * 8,
training=training)
conv4_2 = layers.conv3D_layer_bn(conv4_1,
'conv4_2',
num_filters=n_ch_0 * 8,
training=training)
upconv3 = layers.deconv3D_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.conv3D_layer_bn(concat3,
'conv5_1',
num_filters=n_ch_0 * 4,
training=training)
conv5_2 = layers.conv3D_layer_bn(conv5_1,
'conv5_2',
num_filters=n_ch_0 * 4,
training=training)
upconv2 = layers.deconv3D_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.conv3D_layer_bn(concat2,
'conv6_1',
num_filters=n_ch_0 * 2,
training=training)
conv6_2 = layers.conv3D_layer_bn(conv6_1,
'conv6_2',
num_filters=n_ch_0 * 2,
training=training)
upconv1 = layers.deconv3D_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)
#concat1 = upconv1
conv8_1 = layers.conv3D_layer_bn(concat1,
'conv8_1',
num_filters=n_ch_0,
training=training)
conv8_2 = layers.conv3D_layer(conv8_1,
'conv8_2',
num_filters=1,
activation=tf.identity)
return conv8_2