def __init__(self,
ch=512,
ch_in=512,
w_ch=512,
upsample=True,
enable_blur=False):
super().__init__()
self.upsample = upsample
self.ch = ch
self.ch_in = ch_in
with self.init_scope():
if not upsample:
self.W = chainer.Parameter(shape=(ch_in, 4, 4))
self.W.data[:] = 1 # w_data_tmp
self.b0 = L.Bias(axis=1, shape=(ch, ))
self.b1 = L.Bias(axis=1, shape=(ch, ))
self.n0 = NoiseBlock(ch)
self.n1 = NoiseBlock(ch)
self.s0 = StyleBlock(w_ch, ch)
self.s1 = StyleBlock(w_ch, ch)
self.c0 = EqualizedConv2d(ch_in, ch, 3, 1, 1, nobias=True)
self.c1 = EqualizedConv2d(ch, ch, 3, 1, 1, nobias=True)
self.blur_k = None
self.enable_blur = enable_blur
def __init__(self, ch, dim_z=256):
super(EncoderBlockBase, self).__init__()
with self.init_scope():
self.c0 = EqualizedConv2d(ch, ch, 3, 1, 1)
self.c1 = EqualizedConv2d(ch, ch, 4, 1, 0)
self.l2 = EqualizedLinear(ch, dim_z, gain=1)
self.bn0 = L.BatchNormalization(ch)
self.bn1 = L.BatchNormalization(ch)
def __init__(self, in_ch, out_ch, enable_blur=False):
super(DiscriminatorBlock, self).__init__()
self.in_ch = in_ch
self.out_ch = out_ch
with self.init_scope():
self.c0 = EqualizedConv2d(in_ch, out_ch, 3, 1, 1)
self.c1 = EqualizedConv2d(out_ch, out_ch, 3, 1, 1)
self.blur_k = None
self.enable_blur = enable_blur
def __init__(self, ch=512, enable_blur=False):
super(StyleGenerator, self).__init__()
self.max_stage = 17
with self.init_scope():
self.blocks = chainer.ChainList(
SynthesisBlock(ch, ch, upsample=False), #4
SynthesisBlock(ch, ch, upsample=True, enable_blur=enable_blur), #8
SynthesisBlock(ch, ch, upsample=True, enable_blur=enable_blur), #16
SynthesisBlock(ch, ch, upsample=True, enable_blur=enable_blur), # 32
SynthesisBlock(ch // 2, ch, upsample=True, enable_blur=enable_blur), #64
SynthesisBlock(ch // 4, ch // 2, upsample=True, enable_blur=enable_blur), #128
SynthesisBlock(ch // 8, ch // 4, upsample=True, enable_blur=enable_blur), #256
SynthesisBlock(ch // 16, ch // 8, upsample=True, enable_blur=enable_blur), #512
SynthesisBlock(ch // 32, ch // 16, upsample=True, enable_blur=enable_blur) #1024
)
self.outs = chainer.ChainList(
EqualizedConv2d(ch, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 2, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 4, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 8, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 16, 3, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 32, 3, 1, 1, 0, gain=1)
)
self.n_blocks = len(self.blocks)
self.image_size = 1024
self.enable_blur = enable_blur
def __init__(self, ch=512, enable_blur=False):
super(Discriminator, self).__init__()
self.max_stage = 17
with self.init_scope():
# NOTE: called in reversed order.
self.blocks = chainer.ChainList(
DiscriminatorBlockBase(ch),
DiscriminatorBlock(ch, ch, enable_blur=enable_blur),
DiscriminatorBlock(ch, ch, enable_blur=enable_blur),
DiscriminatorBlock(ch, ch, enable_blur=enable_blur),
DiscriminatorBlock(ch // 2, ch, enable_blur=enable_blur),
DiscriminatorBlock(ch // 4, ch // 2, enable_blur=enable_blur),
DiscriminatorBlock(ch // 8, ch // 4, enable_blur=enable_blur),
DiscriminatorBlock(ch // 16, ch // 8, enable_blur=enable_blur),
DiscriminatorBlock(ch // 32, ch // 16, enable_blur=enable_blur),)
self.ins = chainer.ChainList(
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch // 2, 1, 1, 0),
EqualizedConv2d(3, ch // 4, 1, 1, 0),
EqualizedConv2d(3, ch // 8, 1, 1, 0),
EqualizedConv2d(3, ch // 16, 1, 1, 0),
EqualizedConv2d(3, ch // 32, 1, 1, 0),)
self.enable_blur = enable_blur
def __init__(self,
in_ch=128,
ch=512,
enable_blur=False,
rgbd=False,
use_encoder=False,
use_occupancy_net=False,
initial_depth=None):
super(DCGANGenerator, self).__init__()
self.in_ch = in_ch
self.ch = ch
self.max_stage = 17
self.rgbd = rgbd
self.use_occupancy_net = use_occupancy_net
out_ch = 4 if rgbd else 3
if initial_depth is None:
initial_depth = 1.0
with self.init_scope():
if self.rgbd:
self.linear = EqualizedLinear(in_ch + 9, ch * 4 * 4)
else:
self.linear = EqualizedLinear(in_ch, ch * 4 * 4)
self.blocks = chainer.ChainList(
DCGANBlock(ch, ch, enable_blur=enable_blur), # 8
DCGANBlock(ch, ch, enable_blur=enable_blur), # 16
DCGANBlock(ch, ch, enable_blur=enable_blur), # 32
DCGANBlock(ch // 2, ch, enable_blur=enable_blur), # 64
DCGANBlock(ch // 4, ch // 2, enable_blur=enable_blur), # 128
)
self.outs = chainer.ChainList(
EqualizedConv2d(ch, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 2, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 4, out_ch, 1, 1, 0, gain=1),
)
if use_encoder:
self.enc = Encoder(ch, in_ch, enable_blur=enable_blur)
if use_occupancy_net:
self.occupancy = OccupancyNet(in_ch=in_ch + 3, hidden_ch=32)
# initialize depth weight to 0
for out in self.outs:
out.c.W.array[-1] = 0
out.c.b.array[-1] = math.log(math.e**initial_depth - 1)
self.n_blocks = len(self.blocks)
self.image_size = 128
self.enable_blur = enable_blur
def __init__(self, ch=512, ch_in=512, upsample=True, enable_blur=False):
super().__init__()
self.upsample = upsample
self.ch = ch
self.ch_in = ch_in
with self.init_scope():
self.b0 = L.Bias(axis=1, shape=(ch, ))
self.b1 = L.Bias(axis=1, shape=(ch, ))
self.n0 = NoiseBlock(ch)
self.n1 = NoiseBlock(ch)
self.c0 = EqualizedConv2d(ch_in, ch, 3, 1, 1, nobias=True)
self.c1 = EqualizedConv2d(ch, ch, 3, 1, 1, nobias=True)
self.blur_k = None
self.enable_blur = enable_blur
def __init__(self, ch=512, enable_blur=False, sn=False):
super(Discriminator, self).__init__()
self.max_stage = 17
self.sn = sn
with self.init_scope():
# NOTE: called in reversed order.
self.blocks = chainer.ChainList(
DiscriminatorBlockBase(ch, sn=sn),
DiscriminatorBlock(ch, ch, enable_blur=enable_blur, sn=sn),
DiscriminatorBlock(ch, ch, enable_blur=enable_blur, sn=sn),
DiscriminatorBlock(ch, ch, enable_blur=enable_blur, sn=sn),
DiscriminatorBlock(ch // 2, ch, enable_blur=enable_blur,
sn=sn),
)
if not sn:
self.ins = chainer.ChainList(
EqualizedConv2d(3, ch // 2, 1, 1, 0), )
else:
w = chainer.initializers.GlorotUniform(math.sqrt(2))
self.ins = chainer.ChainList(
L.Convolution2D(3, ch // 2, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()), )
self.enable_blur = enable_blur
def __init__(self,
in_channels,
out_channels,
kernel_size,
bias=True,
padding_layer=None):
'''
:param in_channels: Number of input channels
:param out_channels: Number of output channels
:param kernel_size: Scalar. Spatial dimensions of kernel (only quadratic kernels supported).
:param bias: Whether or not to use bias.
:param padding_layer: Which padding to use. Default is reflection padding.
'''
assert padding_layer is None, "this padding is not supported"
super().__init__()
ka = kernel_size // 2
kb = ka - 1 if kernel_size % 2 == 0 else ka
with self.init_scope():
self.net = chainer.Sequential(
ReflectionPad(pad_width=((ka, kb), (ka, kb))),
EqualizedConv2d(in_channels,
out_channels,
kernel_size,
nobias=not bias,
stride=1))
def __init__(self, ch, out_dim=1, sn=False):
super(DiscriminatorBlockBase, self).__init__()
with self.init_scope():
if not sn:
self.c0 = EqualizedConv2d(ch, ch, 3, 1, 1)
self.c1 = EqualizedConv2d(ch, ch, 4, 1, 0)
self.l2 = EqualizedLinear(ch, out_dim, gain=1)
else:
w = chainer.initializers.Uniform(1)
self.c0 = L.Convolution2D(ch, ch, 3, 1, 1,
initialW=w).add_hook(
SpectralNormalization())
self.c1 = L.Convolution2D(ch, ch, 4, 1, 0,
initialW=w).add_hook(
SpectralNormalization())
self.l2 = L.Linear(ch, out_dim, initialW=w).add_hook(
SpectralNormalization())
def __init__(self, nf0, occnet_nf, frustrum_dims):
super().__init__()
self.occnet_nf = 32 # TODO avoid hard coding
self.frustrum_dims = frustrum_dims
self.frustrum_depth = frustrum_dims[-1]
self.depth_coords = None
with self.init_scope():
self.mlp = chainer.Sequential(
EqualizedConv2d(nf0 * self.frustrum_depth, self.occnet_nf, 1,
1, 0),
L.BatchNormalization(self.occnet_nf),
F.leaky_relu,
EqualizedConv2d(self.occnet_nf, self.occnet_nf),
L.BatchNormalization(self.occnet_nf),
F.leaky_relu,
)
def __init__(self,
in_ch,
out_ch,
enable_blur=False,
sn=False,
res=False,
bn=False):
super(DiscriminatorBlock, self).__init__()
self.in_ch = in_ch
self.out_ch = out_ch
self.res = res
with self.init_scope():
if not sn:
self.c0 = EqualizedConv2d(in_ch, out_ch, 3, 1, 1)
self.c1 = EqualizedConv2d(out_ch, out_ch, 3, 1, 1)
if res:
self.c_sc = EqualizedConv2d(in_ch, out_ch, 3, 1, 1)
else:
w = chainer.initializers.Uniform(1)
self.c0 = L.Convolution2D(in_ch, out_ch, 3, 1, 1,
initialW=w).add_hook(
SpectralNormalization())
self.c1 = L.Convolution2D(out_ch, out_ch, 3, 1, 1,
initialW=w).add_hook(
SpectralNormalization())
if res:
self.c_sc = L.Convolution2D(in_ch,
out_ch,
3,
1,
1,
initialW=w).add_hook(
SpectralNormalization())
if bn:
self.b0 = L.BatchNormalization(out_ch)
self.b1 = L.BatchNormalization(out_ch)
else:
self.b0 = lambda x: x
self.b1 = lambda x: x
self.blur_k = None
self.enable_blur = enable_blur
def __init__(self, ch=512, out_ch=64, enable_blur=False):
super(BackgroundFeatureGenerator, self).__init__()
self.img_size = 64
self.background_depth = 4 # virtual background distance
x_pos, y_pos = np.meshgrid(
np.arange(self.img_size) - self.img_size // 2,
np.arange(self.img_size) - self.img_size // 2)
depth_map = self.background_depth * \
self.img_size * 2 / np.sqrt((self.img_size * 2) ** 2 + x_pos ** 2 + y_pos ** 2)
with self.init_scope():
self.blocks = chainer.ChainList(
SynthesisBlock(ch, ch, ch, upsample=False), # 4
SynthesisBlock(ch,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 8
SynthesisBlock(ch,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 16
SynthesisBlock(ch,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 32
SynthesisBlock(ch // 2,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 64
)
self.conv = EqualizedConv2d(ch // 2, out_ch, 1, 1, 0, gain=1)
self.l1 = EqualizedLinear(ch + 9, ch)
self.l2 = EqualizedLinear(ch, ch)
self.add_persistent("depth_map", depth_map)
self.n_blocks = len(self.blocks)
self.image_size = 64
self.enable_blur = enable_blur
def __init__(self, ch=512, dim_z=256, enable_blur=False, res=True):
super(Encoder, self).__init__()
self.max_stage = 17
with self.init_scope():
# NOTE: called in reversed order.
self.blocks = chainer.ChainList(
EncoderBlockBase(ch, dim_z + 9), # 4
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=False,
res=res,
bn=True), # 8
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=False,
res=res,
bn=True), # 16
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=False,
res=res,
bn=True), # 32
DiscriminatorBlock(ch // 2,
ch,
enable_blur=enable_blur,
sn=False,
res=res,
bn=True), # 64
DiscriminatorBlock(ch // 4,
ch // 2,
enable_blur=enable_blur,
sn=False,
res=res,
bn=True), # 128
)
self.ins = chainer.ChainList(
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch // 2, 1, 1, 0),
EqualizedConv2d(3, ch // 4, 1, 1, 0),
)
self.enable_blur = enable_blur
def __init__(self, w_ch, in_ch, hidden_ch=256):
super(StyleGenerator, self).__init__()
with self.init_scope():
self.c0 = EqualizedConv2d(in_ch, hidden_ch * 2, 4, 2, 1)
self.c1 = EqualizedConv2d(hidden_ch * 2, hidden_ch * 4, 4, 2, 1)
self.c4 = EqualizedConv2d(hidden_ch * 4, hidden_ch * 4, 3, 1, 1)
self.c5 = EqualizedConv2d(hidden_ch * 4, hidden_ch * 2, 3, 1, 1)
self.c6 = EqualizedConv2d(hidden_ch * 2 * 2, hidden_ch, 3, 1, 1)
self.c7 = EqualizedConv2d(hidden_ch + in_ch, 3, 3, 1, 1, gain=0.5)
self.s0 = StyleBlock(w_ch, hidden_ch * 2)
self.s1 = StyleBlock(w_ch, hidden_ch * 4)
self.s4 = StyleBlock(w_ch, hidden_ch * 4)
self.s5 = StyleBlock(w_ch, hidden_ch * 2)
self.s6 = StyleBlock(w_ch, hidden_ch)
def __init__(self, ch):
super(DiscriminatorBlockBase, self).__init__()
with self.init_scope():
self.c0 = EqualizedConv2d(ch, ch, 3, 1, 1)
self.c1 = EqualizedConv2d(ch, ch, 4, 1, 0)
self.l2 = EqualizedLinear(ch, 1, gain=1)
def __init__(self,
in_channels,
out_channels,
prep_conv=True,
middle_channels=None,
use_dropout=False,
dropout_prob=0.1,
norm=L.BatchNormalization):
'''
:param in_channels: Number of input channels
:param out_channels: Number of output channels
:param prep_conv: Whether to have another convolutional layer before the downsampling layer.
:param middle_channels: If prep_conv is true, this sets the number of channels between the prep and downsampling
convs.
:param use_dropout: bool. Whether to use dropout or not.
:param dropout_prob: Float. The dropout probability (if use_dropout is True)
:param norm: Which norm to use. If None, no norm is used. Default is Batchnorm with affinity.
'''
super().__init__()
if middle_channels is None:
middle_channels = in_channels
net = list()
if prep_conv:
net += [
ReflectionPad(pad_width=1),
EqualizedConv2d(in_channels,
middle_channels,
ksize=3,
pad=0,
stride=1,
nobias=False if norm is None else True)
]
if norm is not None:
net += [norm(middle_channels)]
net += [F.leaky_relu]
if use_dropout:
net += [lambda x: F.dropout(x, dropout_prob)]
net += [
ReflectionPad(pad_width=1),
EqualizedConv2d(middle_channels,
out_channels,
ksize=4,
pad=0,
stride=2,
nobias=False if norm is None else True)
]
if norm is not None:
net += [norm(out_channels)]
net += [F.leaky_relu]
if use_dropout:
net += [lambda x: F.dropout(x, dropout_prob)]
with self.init_scope():
self.net = chainer.Sequential(*net)
def __init__(self,
ch=512,
enable_blur=False,
rgbd=False,
rotate_conv_input=False,
use_encoder=False,
use_occupancy_net=False,
initial_depth=1.0):
super(StyleGenerator, self).__init__()
self.max_stage = 17
self.rgbd = rgbd
self.rotate_conv_input = rotate_conv_input
self.use_occupancy_net = use_occupancy_net
out_ch = 4 if rgbd else 3
with self.init_scope():
self.blocks = chainer.ChainList(
SynthesisBlock(ch, ch, ch, upsample=False), # 4
SynthesisBlock(ch,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 8
SynthesisBlock(ch,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 16
SynthesisBlock(ch,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 32
SynthesisBlock(ch // 2,
ch,
ch,
upsample=True,
enable_blur=enable_blur), # 64
SynthesisBlock(ch // 4,
ch // 2,
ch,
upsample=True,
enable_blur=enable_blur), # 128
# SynthesisBlock(ch // 8, ch // 4, ch, upsample=True, enable_blur=enable_blur), # 256
# SynthesisBlock(ch // 16, ch // 8, ch, upsample=True, enable_blur=enable_blur), # 512
# SynthesisBlock(ch // 32, ch // 16, ch, upsample=True, enable_blur=enable_blur) # 1024
)
self.outs = chainer.ChainList(
EqualizedConv2d(ch, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 2, out_ch, 1, 1, 0, gain=1),
EqualizedConv2d(ch // 4, out_ch, 1, 1, 0, gain=1),
# EqualizedConv2d(ch // 8, out_ch, 1, 1, 0, gain=1),
# EqualizedConv2d(ch // 16, out_ch, 1, 1, 0, gain=1),
# EqualizedConv2d(ch // 32, out_ch, 1, 1, 0, gain=1)
)
if self.rgbd:
if self.rotate_conv_input:
self.l1 = EqualizedLinear(9, ch)
else:
self.l1 = EqualizedLinear(ch + 9, ch)
self.l2 = EqualizedLinear(ch, ch)
if rotate_conv_input:
self.rotate = StyleBlock(ch, ch)
if use_encoder:
self.enc = Encoder(ch, ch * 2, enable_blur=enable_blur)
if use_occupancy_net:
self.occupancy = OccupancyNet(in_ch=ch * 2 + 3, hidden_ch=32)
# initialize depth weight to 0
for out in self.outs:
out.c.W.array[-1] = 0
out.c.b.array[-1] = math.log(math.e**initial_depth - 1)
self.n_blocks = len(self.blocks)
self.image_size = 128
self.enable_blur = enable_blur
def __init__(self,
ch=512,
enable_blur=False,
sn=False,
res=False,
num_z=2):
super(DisentangledDiscriminator, self).__init__()
self.max_stage = 17
self.sn = sn
with self.init_scope():
# NOTE: called in reversed order.
self.shared_blocks = chainer.ChainList(
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 16
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 32
DiscriminatorBlock(ch // 2,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 64
DiscriminatorBlock(ch // 4,
ch // 2,
enable_blur=enable_blur,
sn=sn,
res=res), # 128
DiscriminatorBlock(ch // 8,
ch // 4,
enable_blur=enable_blur,
sn=sn,
res=res), # 256
DiscriminatorBlock(ch // 16,
ch // 8,
enable_blur=enable_blur,
sn=sn,
res=res), # 512
DiscriminatorBlock(ch // 32,
ch // 16,
enable_blur=enable_blur,
sn=sn,
res=res),
) # 1024
self.camera_parameter_blocks = chainer.Sequential( # camera parameter variant feature
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 8
DiscriminatorBlockBase(
ch, sn=sn,
out_dim=9), # 4 # Euler angles, translation vector
)
self.z_regression_blocks = chainer.Sequential( # camera parameter invariant feature
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 8
DiscriminatorBlockBase(ch, sn=sn, out_dim=ch * num_z), # 4
)
self.discriminator_blocks = chainer.Sequential( # camera parameter variant feature
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 8
DiscriminatorBlockBase(
ch, sn=sn), # 4 # Euler angles*2, translation vector
)
if not sn:
self.ins = chainer.ChainList(
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch // 2, 1, 1, 0),
EqualizedConv2d(3, ch // 4, 1, 1, 0),
EqualizedConv2d(3, ch // 8, 1, 1, 0),
EqualizedConv2d(3, ch // 16, 1, 1, 0),
EqualizedConv2d(3, ch // 32, 1, 1, 0),
)
else:
w = chainer.initializers.Uniform(1)
self.ins = chainer.ChainList(
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch // 2, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch // 4, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch // 8, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch // 16, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch // 32, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
)
self.camera_param_discriminator = CameraParamDiscriminator()
self.enable_blur = enable_blur
def __init__(self,
ch=512,
out_dim=1,
enable_blur=False,
sn=False,
res=False):
super(Discriminator, self).__init__()
self.max_stage = 17
self.sn = sn
with self.init_scope():
# NOTE: called in reversed order.
self.blocks = chainer.ChainList(
DiscriminatorBlockBase(ch, out_dim, sn=sn), # 4
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 8
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 16
DiscriminatorBlock(ch,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 32
DiscriminatorBlock(ch // 2,
ch,
enable_blur=enable_blur,
sn=sn,
res=res), # 64
DiscriminatorBlock(ch // 4,
ch // 2,
enable_blur=enable_blur,
sn=sn,
res=res), # 128
)
if not sn:
self.ins = chainer.ChainList(
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch, 1, 1, 0),
EqualizedConv2d(3, ch // 2, 1, 1, 0),
EqualizedConv2d(3, ch // 4, 1, 1, 0),
)
else:
w = chainer.initializers.Uniform(1)
self.ins = chainer.ChainList(
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch // 2, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
L.Convolution2D(3, ch // 4, 1, 1, 0, initialW=w).add_hook(
SpectralNormalization()),
)
self.enable_blur = enable_blur
