def __init__(self, x_channels=1, y_channels=3, ch=64, activation=F.relu):
super().__init__()
self.activation = activation
initializer = chainer.initializers.GlorotUniform()
with self.init_scope():
self.block0_0 = OptimizedBlock(x_channels, ch // 2)
self.block0_1 = OptimizedBlock(y_channels, ch // 2)
self.block1 = Block(ch,
ch * 2,
activation=activation,
downsample=True)
self.block2 = Block(ch * 2,
ch * 4,
activation=activation,
downsample=True)
self.block3 = Block(ch * 4,
ch * 8,
activation=activation,
downsample=True)
self.block4 = Block(ch * 8,
ch * 16,
activation=activation,
downsample=True)
self.block5 = Block(ch * 16,
ch * 16,
activation=activation,
downsample=False)
self.l6 = Linear(ch * 16, 1, initialW=initializer)
def __init__(self, ch=64, n_classes=0, activation=F.relu):
super(SNResNetProjectionDiscriminator, self).__init__()
self.activation = activation
initializer = chainer.initializers.GlorotUniform()
with self.init_scope():
self.block1 = OptimizedBlock(3, ch)
self.block2 = Block(ch,
ch * 2,
activation=activation,
downsample=True)
self.block3 = Block(ch * 2,
ch * 4,
activation=activation,
downsample=True)
self.block4 = Block(ch * 4,
ch * 8,
activation=activation,
downsample=True)
self.block5 = Block(ch * 8,
ch * 16,
activation=activation,
downsample=True)
self.block6 = Block(ch * 16,
ch * 16,
activation=activation,
downsample=False)
self.l7 = SNLinear(ch * 16, 1, initialW=initializer)
if n_classes > 0:
self.l_y = SNEmbedID(n_classes, ch * 16, initialW=initializer)
def __init__(self, ch=64, ch_in=3, n_classes=0, activation=F.relu):
super(SNResNetProjectionDiscriminator, self).__init__()
self.activation = activation
self.block1 = OptimizedBlock(ch_in, ch)
self.block2 = Block(ch, ch * 2, activation=activation, downsample=True)
self.block3 = Block(ch * 2,
ch * 4,
activation=activation,
downsample=True)
self.block4 = Block(ch * 4,
ch * 8,
activation=activation,
downsample=True)
self.block5 = Block(ch * 8,
ch * 16,
activation=activation,
downsample=True)
self.block6 = Block(ch * 16,
ch * 16,
activation=activation,
downsample=False)
self.l7 = SNLinear(ch * 16, 1)
if n_classes > 0:
self.l_y = SNEmbedding(n_classes, ch * 16)
def __init__(self, ch=128, n_classes=0, activation=F.relu):
super(SNResNetProjectionDiscriminator, self).__init__()
self.activation = activation
with self.init_scope():
self.block1 = OptimizedBlock(3, ch)
self.block2 = Block(ch, ch, activation=activation, downsample=True)
self.block3 = Block(ch,
ch,
activation=activation,
downsample=False)
self.block4 = Block(ch,
ch,
activation=activation,
downsample=False)
self.l5 = SNLinear(ch,
1,
initialW=chainer.initializers.GlorotUniform(),
nobias=True)
self.l5.u = np.ones((1, 1)).astype(np.float32)
if n_classes > 0:
self.l_y = SNEmbedID(
n_classes,
ch,
initialW=chainer.initializers.GlorotUniform())
seed_weights(self)
def __init__(self,
init_ch=1,
ch=6,
activation=F.relu,
dim_z=4,
bottom_size=32):
super(Discriminator, self).__init__()
self.activation = activation
initializer = chainer.initializers.GlorotUniform()
with self.init_scope():
# 1x32x32
self.phi_block1 = OptimizedBlock(init_ch, ch)
self.phi_block2 = Block(ch,
ch * 2,
activation=activation,
downsample=True)
self.phi_block3 = Block(ch * 2,
ch * 2,
activation=activation,
downsample=True)
# ch*2x8x8 => basically 6*2x8*8 = 768
self.enc1 = Block1(init_ch,
ch,
activation=activation,
is_shortcut=True,
dim_z=dim_z)
self.enc2 = Block1(ch,
ch * 2,
activation=activation,
is_shortcut=True,
dim_z=dim_z)
self.enc3 = Block1(ch * 2,
ch * 2,
activation=activation,
is_shortcut=True,
dim_z=dim_z)
self.linear = L.Linear(
ch * 2 * (bottom_size * bottom_size),
ch * 2 * (bottom_size // 8 * bottom_size // 8))
self.l8 = SNLinear(ch * 2 * (bottom_size // 8) *
(bottom_size // 8),
1,
initialW=initializer)
def __init__(self, ch=128, n_classes=0, activation=F.relu):
super(SRDiscriminator, self).__init__()
self.activation = activation
with self.init_scope():
self.block1 = OptimizedBlock(3, ch)
self.block2 = Block(ch, ch, activation=activation, downsample=True)
self.block3 = Block(ch,
ch,
activation=activation,
downsample=False)
self.block4 = Block(ch,
ch,
activation=activation,
downsample=False)
self.l5 = SNLinear(ch,
1,
initialW=chainer.initializers.GlorotUniform(),
nobias=True)
if n_classes > 0:
self.l_y = SNEmbedID(
n_classes,
ch,
initialW=chainer.initializers.GlorotUniform())