def __init__(self):
super(Branch2, self).__init__()
self.conv_input = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=1,
padding=1,
bias=False)
# self.conv_input2 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, stride=1, padding=1, bias=False)
self.relu = nn.PReLU()
self.convt_F11 = CARB(64)
self.convt_F12 = CARB(64)
self.convt_F13 = CARB(64)
# style encode
self.s_conv1 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=1,
padding=1,
bias=False)
self.s_conv2 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=2,
padding=1,
bias=False)
self.s_conv3 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=2,
padding=1,
bias=False)
#-------------
self.u1 = upsample_block(64, 256)
self.convt_shape1 = nn.Conv2d(in_channels=64,
out_channels=3,
kernel_size=3,
stride=1,
padding=1,
bias=False)
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
if m.bias is not None:
m.bias.data.zero_()
def __init__(self):
super(Branch6, self).__init__()
self.conv_input = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=1,
padding=1,
bias=False)
# self.conv_input2 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, stride=1, padding=1, bias=False)
self.relu = nn.PReLU()
self.convt_F11 = CARB(64)
self.convt_F12 = CARB(64)
self.convt_F13 = CARB(64)
self.convt_F14 = CARB(64)
self.convt_F15 = CARB(64)
self.convt_F16 = CARB(64)
self.convt_F17 = CARB(64)
self.convt_F18 = CARB(64)
self.s_conv1 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=1,
padding=1,
bias=False)
self.s_conv2 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=2,
padding=1,
bias=False)
self.s_conv3 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=1,
padding=1,
bias=False)
self.s_conv4 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=2,
padding=1,
bias=False)
self.s_conv5 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=1,
padding=1,
bias=False)
self.s_conv6 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=2,
padding=1,
bias=False)
self.s_conv7 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=1,
padding=1,
bias=False)
self.s_conv8 = nn.Conv2d(in_channels=64,
out_channels=64,
kernel_size=3,
stride=2,
padding=1,
bias=False)
self.u1 = upsample_block(64, 256)
self.u2 = upsample_block(64, 256)
self.u3 = upsample_block(64, 256)
self.u4 = upsample_block(64, 256)
self.u5 = upsample_block(64, 256)
self.convt_shape1 = nn.Conv2d(in_channels=64,
out_channels=3,
kernel_size=3,
stride=1,
padding=1,
bias=False)
self.non_local = Nonlocal_CA(in_feat=64,
inter_feat=64 // 8,
reduction=8,
sub_sample=False,
bn_layer=False)
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
if m.bias is not None:
m.bias.data.zero_()