File size: 9,043 Bytes
b77fd1a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
import torch
import torch.nn as nn
import torch.nn.functional as F

class AttentionBlock(nn.Module):
    def __init__(self, in_ch):
        super().__init__()
        self.group_norm = nn.GroupNorm(32, in_ch)
        self.proj_q = nn.Conv2d(in_ch, in_ch, 1, stride=1, padding=0)
        self.proj_k = nn.Conv2d(in_ch, in_ch, 1, stride=1, padding=0)
        self.proj_v = nn.Conv2d(in_ch, in_ch, 1, stride=1, padding=0)
        self.proj = nn.Conv2d(in_ch, in_ch, 1, stride=1, padding=0)

    def forward(self, x):
        B, C, H, W = x.shape
        h = self.group_norm(x)
        q = self.proj_q(h)
        k = self.proj_k(h)
        v = self.proj_v(h)

        q = q.permute(0, 2, 3, 1).view(B, H * W, C)
        k = k.view(B, C, H * W)
        w = torch.bmm(q, k) * (int(C) ** (-0.5))

        w = F.softmax(w, dim=-1)

        v = v.permute(0, 2, 3, 1).view(B, H * W, C)
        h = torch.bmm(w, v)
        assert list(h.shape) == [B, H * W, C]
        h = h.view(B, H, W, C).permute(0, 3, 1, 2)
        h = self.proj(h)

        return x + h


class ResidualBlock(nn.Module):
    def __init__(self,
                 in_channels: int,
                 out_channels: int,
                 dropout: float,
                 n_groups: int = 32,
                 has_attn: bool = False):
        super().__init__()

        self.norm1 = nn.GroupNorm(n_groups, in_channels)
        self.act1 = nn.SiLU()
        self.conv1 = nn.Conv2d(in_channels, out_channels,
                               kernel_size=(3, 3), padding=(1, 1))

        self.norm2 = nn.GroupNorm(n_groups, out_channels)
        self.act2 = nn.SiLU()
        self.conv2 = nn.Conv2d(out_channels, out_channels,
                               kernel_size=(3, 3), padding=(1, 1))

        if in_channels != out_channels:
            self.shortcut = nn.Conv2d(
                in_channels, out_channels, kernel_size=(1, 1))
        else:
            self.shortcut = nn.Identity()

        if has_attn:
            self.attn = AttentionBlock(out_channels)
        else:
            self.attn = nn.Identity()

        self.dropout = nn.Dropout(dropout)

    def forward(self, x: torch.Tensor):
        h = self.conv1(self.act1(self.norm1(x)))
        h = self.conv2(self.dropout(self.act2(self.norm2(h))))
        return self.attn(h + self.shortcut(x))


class DownBlock(nn.Module):
    def __init__(self, in_channels: int, out_channels: int, has_attn: bool, dropout: int):
        super().__init__()
        self.res = ResidualBlock(
            in_channels, out_channels, dropout=dropout, has_attn=has_attn)

    def forward(self, x: torch.Tensor):
        return self.res(x)


class UpBlock(nn.Module):
    def __init__(self, in_channels: int, out_channels: int, has_attn: bool, dropout: int):
        super().__init__()
        self.res = ResidualBlock(
            in_channels, out_channels, dropout=dropout, has_attn=has_attn)

    def forward(self, x: torch.Tensor):
        return self.res(x)


class MiddleBlock(nn.Module):
    def __init__(self, n_channels: int, dropout: int):
        super().__init__()
        self.res1 = ResidualBlock(
            n_channels, n_channels, dropout=dropout, has_attn=True)
        self.res2 = ResidualBlock(n_channels, n_channels, dropout=dropout)

    def forward(self, x: torch.Tensor):
        x = self.res1(x)
        x = self.res2(x)
        return x


class Downsample(nn.Module):
    def __init__(self, n_channels):
        super().__init__()
        self.conv = nn.Conv2d(n_channels, n_channels,
                              kernel_size=3, stride=2, padding=1)

    def forward(self, x: torch.Tensor):
        return self.conv(x)


class Upsample(nn.Module):
    def __init__(self, n_channels):
        super().__init__()
        self.convT = nn.ConvTranspose2d(
            n_channels, n_channels, kernel_size=3, stride=2, padding=1, output_padding=1)
        self.conv = nn.Conv2d(n_channels, n_channels,
                              kernel_size=3, stride=1, padding=1)

    def forward(self, x: torch.Tensor):
        # Bx, Cx, Hx, Wx = x.size()
        # x = F.interpolate(x, size=(2*Hx, 2*Wx), mode='bicubic', align_corners=False)
        return self.conv(self.convT(x))


class MeanShift(nn.Conv2d):
    def __init__(
            self, rgb_range,
            rgb_mean=(0.4488, 0.4371, 0.4040), rgb_std=(1.0, 1.0, 1.0), sign=-1):

        super(MeanShift, self).__init__(3, 3, kernel_size=1)
        std = torch.Tensor(rgb_std)
        self.weight.data = torch.eye(3).view(3, 3, 1, 1) / std.view(3, 1, 1, 1)
        self.bias.data = sign * rgb_range * torch.Tensor(rgb_mean) / std
        for p in self.parameters():
            p.requires_grad = False


class UNET(nn.Module):
    def __init__(self,
                 in_channels: int = 3,
                 out_channels: int = 3,
                 n_features: int = 64,
                 dropout: int = 0.1,
                 block_out_channels=[64, 128, 128, 256],
                 layers_per_block=4,
                 is_attn_layers=(False, False, True, False),
                 ):
        super().__init__()

        self.in_channels = in_channels
        self.out_channels = out_channels
        self.n_features = n_features
        self.dropout = dropout
        self.block_out_channels = block_out_channels
        self.layers_per_block = layers_per_block
        self.is_attn_layers = is_attn_layers

        self.sub_mean = MeanShift(255)
        self.add_mean = MeanShift(255, sign=1)

        self.shallow_feature_extraction = nn.Conv2d(
            in_channels, n_features, kernel_size=3, padding=1)
        self.image_rescontruction = nn.Conv2d(
            n_features, in_channels, kernel_size=3, padding=1)

        self.left_model = self.left_unet()
        self.middle_model = MiddleBlock(
            block_out_channels[-1], dropout=self.dropout)
        self.right_model = self.right_unet()

    def left_unet(self):
        left_model = []

        in_channel = out_channel = self.n_features
        for i in range(len(self.block_out_channels)):
            out_channel = self.block_out_channels[i]

            down_block = [DownBlock(in_channel, out_channel, dropout=self.dropout, has_attn=self.is_attn_layers[i])] \
                + [DownBlock(out_channel, out_channel, dropout=self.dropout,
                             has_attn=self.is_attn_layers[i])] * (self.layers_per_block - 1)
            in_channel = out_channel
            left_model.append(nn.Sequential(*down_block))
            if i < len(self.block_out_channels):
                left_model.append(Downsample(out_channel))

        return nn.ModuleList(left_model)

    def right_unet(self):
        right_unet = []

        in_channel = out_channel = self.block_out_channels[-1]
        for i in reversed(range(len(self.block_out_channels))):

            out_channel = self.block_out_channels[i]

            up_block = [UpBlock(in_channel, out_channel, dropout=self.dropout, has_attn=self.is_attn_layers[i - 1])] \
                + [UpBlock(out_channel, out_channel, dropout=self.dropout, has_attn=self.is_attn_layers[i - 1])
                   ] * (self.layers_per_block - 1)

            in_channel = out_channel * 2
            right_unet.append(nn.Sequential(*up_block))
            right_unet.append(Upsample(out_channel))

        in_channel, out_channel = self.block_out_channels[0] * \
            2, self.n_features
        up_block = [UpBlock(in_channel, out_channel, dropout=self.dropout, has_attn=self.is_attn_layers[0])] \
            + [UpBlock(out_channel, out_channel, dropout=self.dropout, has_attn=self.is_attn_layers[0])
               ] * (self.layers_per_block - 1)
        right_unet.append(nn.Sequential(*up_block))
        return nn.ModuleList(right_unet)

    def forward(self, x):
        x = x * 255
        x = self.sub_mean(x)

        feature_maps = self.shallow_feature_extraction(x)
        feature_x = [feature_maps]
        # print(feature_maps.shape)
        feature_block = feature_maps
        for block in self.left_model:
            feature_block = block(feature_block)
            if not isinstance(block, Downsample):
                # print(feature_block.shape)
                feature_x.append(feature_block)

        bottleneck = self.middle_model(feature_block)

        feature_x.reverse()
        # print('Middle::: ', feature_maps.shape)

        recover = bottleneck
        d = 0
        for block in self.right_model:
            if isinstance(block, Upsample):
                # print('UP-CAT::: ', recover.shape)
                recover = block(recover)
                # print('UP-CAT-END::: ', recover.shape, feature_x[d].shape)
                recover = torch.cat([recover, feature_x[d]], 1)
                # print('UP-CAT-END::: ', recover.shape, feature_x[d].shape)
                d += 1
            else:
                recover = block(recover)
                # print('UP-RES::: ', recover.shape)

        recover = self.image_rescontruction(recover)
        recover = self.add_mean(recover) / 255
        return recover