def generator(self, z, y, is_training=True, reuse=False):
# Network Architecture is exactly same as in infoGAN (https://arxiv.org/abs/1606.03657)
# Architecture : FC1024_BR-FC7x7x128_BR-(64)4dc2s_BR-(1)4dc2s_S
with tf.variable_scope('generator', reuse=reuse):
# merge noise and condition label
z = concat([z, y], axis=1)
net = tf.nn.relu(
batch_norm(linear(z, 1024, scope='g_fc1'),
is_training=is_training,
scope='g_bn1'))
net = tf.nn.relu(
batch_norm(linear(net, 128 * 7 * 7, scope='g_fc2'),
is_training=is_training,
scope='g_bn2'))
net = tf.reshape(net, [self.batch_size, 7, 7, 128])
net = tf.nn.relu(
batch_norm(deconv2d(net, [self.batch_size, 14, 14, 64], (4, 4),
(2, 2),
name='g_dc3'),
is_training=is_training,
scope='g_bn3'))
out = tf.nn.sigmoid(
deconv2d(net, [self.batch_size, 28, 28, 1], (4, 4), (2, 2),
name='g_dc4'))
return out
def classifier(self, x, is_training=True, reuse=False):
# Network Architecture is exactly same as in infoGAN (https://arxiv.org/abs/1606.03657)
# Architecture : (64)5c2s-(128)5c2s_BL-FC1024_BL-FC128_BL-FC12S’
# All layers except the last two layers are shared by discriminator
with tf.variable_scope("classifier", reuse=reuse):
net = lrelu(
batch_norm(linear(x, 128, scope='c_fc1'),
is_training=is_training,
scope='c_bn1'))
out_logit = linear(net, self.y_dim, scope='c_fc2')
out = tf.nn.softmax(out_logit)
return out, out_logit
def discriminator(self, x, is_training=True, reuse=False):
# Network Architecture is exactly same as in infoGAN (https://arxiv.org/abs/1606.03657)
# Architecture : (64)4c2s-(128)4c2s_BL-FC1024_BL-FC1_S
with tf.variable_scope('discriminator', reuse=reuse):
net = lrelu(conv2d(x, 64, (4, 4), (2, 2), name='d_conv1'))
net = lrelu(
batch_norm(conv2d(net, 128, (4, 4), (2, 2), name='d_conv2'),
is_training=is_training,
scope='d_bn2'))
net = tf.reshape(net, [self.batch_size, -1])
net = lrelu(
batch_norm(linear(net, 1024, scope='d_fc3'),
is_training=is_training,
scope='d_bn3'))
out_logit = linear(net, 1, scope='d_fc4')
out = tf.nn.sigmoid(out_logit)
return out, out_logit, net