File size: 5,820 Bytes
4fcebd2 7cc5832 e1b9be9 7cc5832 4fcebd2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 |
import torch
import torch.nn as nn
import torch.nn.functional as F
# Modules from model
try:
from archs.arch_util import LayerNorm2d
import archs.arch_util as arch_util
except:
from arch_util import LayerNorm2d
import arch_util as arch_util
# Process Block 4 en SFNet y 5 bloques en AmpNet, con el spatial block aplicado en AmpNet (frequency stage)
# tal y como lo tienen ellos en su github (aunque en el paper es al revés) y no lo aplican el space stage
class SimpleGate(nn.Module):
def forward(self, x):
x1, x2 = x.chunk(2, dim=1)
return x1 * x2
class SpaBlock(nn.Module):
def __init__(self, nc, DW_Expand = 2, FFN_Expand=2, drop_out_rate=0.):
super(SpaBlock, self).__init__()
dw_channel = nc * DW_Expand
self.conv1 = nn.Conv2d(in_channels=nc, out_channels=dw_channel, kernel_size=1, padding=0, stride=1, groups=1, bias=True)
self.conv2 = nn.Conv2d(in_channels=dw_channel, out_channels=dw_channel, kernel_size=3, padding=1, stride=1, groups=dw_channel,
bias=True) # the dconv
self.conv3 = nn.Conv2d(in_channels=dw_channel // 2, out_channels=nc, kernel_size=1, padding=0, stride=1, groups=1, bias=True)
# Simplified Channel Attention
self.sca = nn.Sequential(
nn.AdaptiveAvgPool2d(1),
nn.Conv2d(in_channels=dw_channel // 2, out_channels=dw_channel // 2, kernel_size=1, padding=0, stride=1,
groups=1, bias=True),
)
# SimpleGate
self.sg = SimpleGate()
ffn_channel = FFN_Expand * nc
self.conv4 = nn.Conv2d(in_channels=nc, out_channels=ffn_channel, kernel_size=1, padding=0, stride=1, groups=1, bias=True)
self.conv5 = nn.Conv2d(in_channels=ffn_channel // 2, out_channels=nc, kernel_size=1, padding=0, stride=1, groups=1, bias=True)
self.norm1 = LayerNorm2d(nc)
self.norm2 = LayerNorm2d(nc)
self.dropout1 = nn.Dropout(drop_out_rate) if drop_out_rate > 0. else nn.Identity()
self.dropout2 = nn.Dropout(drop_out_rate) if drop_out_rate > 0. else nn.Identity()
self.beta = nn.Parameter(torch.zeros((1, nc, 1, 1)), requires_grad=True)
self.gamma = nn.Parameter(torch.zeros((1, nc, 1, 1)), requires_grad=True)
def forward(self, x):
x = self.norm1(x) # size [B, C, H, W]
x = self.conv1(x) # size [B, 2*C, H, W]
x = self.conv2(x) # size [B, 2*C, H, W]
x = self.sg(x) # size [B, C, H, W]
x = x * self.sca(x) # size [B, C, H, W]
x = self.conv3(x) # size [B, C, H, W]
x = self.dropout1(x)
y = x + x * self.beta # size [B, C, H, W]
x = self.conv4(self.norm2(y)) # size [B, 2*C, H, W]
x = self.sg(x) # size [B, C, H, W]
x = self.conv5(x) # size [B, C, H, W]
x = self.dropout2(x)
return y + x * self.gamma
class FreBlock(nn.Module):
def __init__(self, nc):
super(FreBlock, self).__init__()
self.fpre = nn.Conv2d(nc, nc, 1, 1, 0)
self.process1 = nn.Sequential(
nn.Conv2d(nc, nc, 1, 1, 0),
nn.LeakyReLU(0.1, inplace=True),
nn.Conv2d(nc, nc, 1, 1, 0))
self.process2 = nn.Sequential(
nn.Conv2d(nc, nc, 1, 1, 0),
nn.LeakyReLU(0.1, inplace=True),
nn.Conv2d(nc, nc, 1, 1, 0))
def forward(self, x):
_, _, H, W = x.shape
x_freq = torch.fft.rfft2(self.fpre(x), norm='backward')
mag = torch.abs(x_freq)
pha = torch.angle(x_freq)
mag = self.process1(mag)
pha = self.process2(pha)
real = mag * torch.cos(pha)
imag = mag * torch.sin(pha)
x_out = torch.complex(real, imag)
x_out = torch.fft.irfft2(x_out, s=(H, W), norm='backward')
return x_out+x
class ProcessBlock(nn.Module):
def __init__(self, in_nc, spatial = True):
super(ProcessBlock,self).__init__()
self.spatial = spatial
self.spatial_process = SpaBlock(in_nc) if spatial else nn.Identity()
self.frequency_process = FreBlock(in_nc)
self.cat = nn.Conv2d(2*in_nc,in_nc,1,1,0) if spatial else nn.Conv2d(in_nc,in_nc,1,1,0)
def forward(self, x):
xori = x
x_freq = self.frequency_process(x)
x_spatial = self.spatial_process(x)
xcat = torch.cat([x_spatial,x_freq],1)
x_out = self.cat(xcat) if self.spatial else self.cat(x_freq)
return x_out+xori
class SFNet(nn.Module):
def __init__(self, nc,n=5):
super(SFNet,self).__init__()
self.list_block = list()
for index in range(n):
self.list_block.append(ProcessBlock(nc,spatial=False))
self.block = nn.Sequential(*self.list_block)
def forward(self, x):
x_ori = x
x_out = self.block(x_ori)
xout = x_ori + x_out
return xout
class AmplitudeNet_skip(nn.Module):
def __init__(self, nc,n=1):
super(AmplitudeNet_skip,self).__init__()
self.conv1 = nn.Sequential(
nn.Conv2d(3, nc, 1, 1, 0),
ProcessBlock(nc),
)
self.conv2 = ProcessBlock(nc)
self.conv3 = ProcessBlock(nc)
self.conv4 = nn.Sequential(
ProcessBlock(nc * 2),
nn.Conv2d(nc * 2, nc, 1, 1, 0),
)
self.conv5 = nn.Sequential(
ProcessBlock(nc * 2),
nn.Conv2d(nc * 2, nc, 1, 1, 0),
)
self.convout = nn.Sequential(
ProcessBlock(nc * 2),
nn.Conv2d(nc * 2, 3, 1, 1, 0),
)
def forward(self, x):
x1 = self.conv1(x)
x2 = self.conv2(x1)
x3 = self.conv3(x2)
x4 = self.conv5(torch.cat((x2, x3), dim=1))
xout = self.convout(torch.cat((x1, x4), dim=1))
return xout
|