def encoder(x, y, phase, scope='enc', reuse=None, internal_update=False):
with tf.variable_scope(scope, reuse=reuse):
with arg_scope([conv2d, dense], bn=True, phase=phase, activation=leaky_relu), \
arg_scope([batch_norm], internal_update=internal_update):
# Ignore y
x = conv2d(x, 64, 3, 2)
x = conv2d(x, 128, 3, 2)
x = conv2d(x, 256, 3, 2)
x = dense(x, 1024)
# Autoregression (4 steps)
ms = []
vs = []
zs = [x]
for i in xrange(5):
h = tf.concat(zs, axis=-1)
h = dense(h, 100)
m = dense(h, 20, activation=None)
v = dense(h, 20, activation=tf.nn.softplus) + 1e-5
z = gaussian_sample(m, v)
ms += [m]
vs += [v]
zs += [z]
m = tf.concat(ms, 1)
v = tf.concat(vs, 1)
z = tf.concat(zs[1:], 1)
return z, (m, v)
def encoder(x, phase, reuse=None):
with tf.variable_scope('enc', reuse=reuse):
with arg_scope([dense], bn=True, phase=phase, activation=leaky_relu), \
arg_scope([conv2d], bn=True, phase=phase, activation=leaky_relu):
if x._shape_as_list()[-1] == 1:
x = tf.image.grayscale_to_rgb(x)
x = conv2d(x, 64, 3, 2)
x = conv2d(x, 128, 3, 2)
x = conv2d(x, 256, 3, 2)
x = dense(x, 128)
return x
def classifier(x, phase, reuse=None):
with tf.variable_scope('class', reuse=reuse):
with arg_scope([conv2d, dense],
bn=True,
phase=phase,
activation=tf.nn.relu):
for i in range(4):
x = conv2d(x, 64 + 64 * i, 3, 2)
x = conv2d(x, 64 + 64 * i, 3, 1)
x = dense(x, 500)
x = dense(x, 10, activation=None)
return x
def encoder(x, y, phase, reuse=None):
with tf.variable_scope('enc', reuse=reuse):
with arg_scope([conv2d, dense], bn=True, phase=phase, activation=leaky_relu), \
arg_scope([noise], phase=phase):
# Ignore y
x = conv2d(x, 64, 3, 2, bn=False)
x = conv2d(x, 128, 3, 2)
x = conv2d(x, 256, 3, 2)
x = dense(x, 1024)
m = dense(x, 100, activation=None)
v = dense(x, 100, activation=tf.nn.softplus) + 1e-5
return (m, v)
def discriminator(x, phase, reuse=None, depth=1):
with tf.variable_scope('disc', reuse=reuse):
with arg_scope([conv2d, dense], bn=True, phase=phase, activation=lrelu), \
arg_scope([noise], phase=phase):
x = dropout(x, rate=0.2, training=phase)
x = conv2d(x, 64, 3, 2, bn=False)
x = dropout(x, training=phase)
x = conv2d(x, 128, 3, 2)
x = dropout(x, training=phase)
x = conv2d(x, 256, 3, 2)
x = dropout(x, training=phase)
x = dense(x, 1024)
x = dense(x, depth, activation=None, bn=False)
return x
def encoder(x, y, phase, scope='enc', reuse=None, internal_update=False):
with tf.variable_scope(scope, reuse=reuse):
with arg_scope([conv2d, dense], bn=True, phase=phase, activation=leaky_relu), \
arg_scope([batch_norm], internal_update=internal_update):
# Ignore y
x = conv2d(x, 64, 3, 2)
x = conv2d(x, 128, 3, 2)
x = conv2d(x, 256, 3, 2)
x = dense(x, 1024)
m = dense(x, 100, activation=None)
v = dense(x, 100, activation=tf.nn.softplus) + 1e-5
z = gaussian_sample(m, v)
return z, (m, v)
def encoder(x, phase, reuse=None):
with tf.variable_scope('enc', reuse=reuse):
with arg_scope([dense], bn=True, phase=phase, activation=leaky_relu), \
arg_scope([conv2d], bn=True, phase=phase, activation=leaky_relu):
if x._shape_as_list()[-1] == 1:
x = tf.image.grayscale_to_rgb(x)
if args.dense:
# Level 0: 32
l0 = x
# Level 1: 32 -> 16, 8, 4
a1 = conv2d(l0, 64, 3, 2)
a2 = conv2d(l0, 64, 5, 4)
a3 = conv2d(l0, 64, 9, 8)
l1 = a1
# Level 2: 16 -> 8, 4
b2 = conv2d(l1, 64, 3, 2)
b3 = conv2d(l1, 64, 5, 4)
l2 = tf.concat([a2, b2], -1)
# Level 3: 8 -> 4
c3 = conv2d(l2, 64, 3, 2)
l3 = tf.concat([a3, b3, c3], -1)
# Level 4: Dense
x = dense(l3, 128)
else:
x = conv2d(x, 64, 3, 2)
x = conv2d(x, 128, 3, 2)
x = conv2d(x, 256, 3, 2)
x = dense(x, 128)
return x
