| 80 | |
| 81 | |
| 82 | class DocNC(nn.Module): |
| 83 | def __init__(self): |
| 84 | super(DocNC, self).__init__() |
| 85 | |
| 86 | # device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu') |
| 87 | # if IsING == True: |
| 88 | # # self.conv1_1 = ING(3,32) |
| 89 | # self.conv1_1 = nn.Sequential(nn.Conv2d(3, 32, kernel_size=3, stride=1, padding=1), |
| 90 | # nn.InstanceNorm2d(32, affine=True)) |
| 91 | # else: |
| 92 | self.conv0 = nn.Conv2d(3, 3, kernel_size=3, stride=1, padding=1) |
| 93 | |
| 94 | self.conv1_1 = AttING(3,32) |
| 95 | # self.conv1_1 = nn.Conv2d(3, 32, kernel_size=3, stride=1, padding=1) |
| 96 | self.conv1_2 = nn.Conv2d(32, 32, kernel_size=3, stride=1, padding=1) |
| 97 | self.pool1 = nn.MaxPool2d(kernel_size=2) |
| 98 | |
| 99 | self.conv2_1 = nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1) |
| 100 | self.conv2_2 = nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1) |
| 101 | self.pool2 = nn.MaxPool2d(kernel_size=2) |
| 102 | |
| 103 | self.conv3_1 = nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1) |
| 104 | self.conv3_2 = nn.Conv2d(128, 128, kernel_size=3, stride=1, padding=1) |
| 105 | self.pool3 = nn.MaxPool2d(kernel_size=2) |
| 106 | |
| 107 | self.conv4_1 = nn.Conv2d(128, 256, kernel_size=3, stride=1, padding=1) |
| 108 | self.conv4_2 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1) |
| 109 | self.pool4 = nn.MaxPool2d(kernel_size=2) |
| 110 | |
| 111 | self.conv5_1 = nn.Conv2d(256, 512, kernel_size=3, stride=1, padding=1) |
| 112 | self.conv5_2 = nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1) |
| 113 | |
| 114 | self.upv6 = nn.ConvTranspose2d(512, 256, 2, stride=2) |
| 115 | self.conv6_1 = nn.Conv2d(512, 256, kernel_size=3, stride=1, padding=1) |
| 116 | self.conv6_2 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1) |
| 117 | |
| 118 | self.upv7 = nn.ConvTranspose2d(256, 128, 2, stride=2) |
| 119 | self.conv7_1 = nn.Conv2d(256, 128, kernel_size=3, stride=1, padding=1) |
| 120 | self.conv7_2 = nn.Conv2d(128, 128, kernel_size=3, stride=1, padding=1) |
| 121 | |
| 122 | self.upv8 = nn.ConvTranspose2d(128, 64, 2, stride=2) |
| 123 | self.conv8_1 = nn.Conv2d(128, 64, kernel_size=3, stride=1, padding=1) |
| 124 | self.conv8_2 = nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1) |
| 125 | |
| 126 | self.upv9 = nn.ConvTranspose2d(64, 32, 2, stride=2) |
| 127 | self.conv9_1 = nn.Conv2d(64, 32, kernel_size=3, stride=1, padding=1) |
| 128 | self.conv9_2 = nn.Conv2d(32, 32, kernel_size=3, stride=1, padding=1) |
| 129 | |
| 130 | self.conv10_1 = nn.Conv2d(32, 3, kernel_size=1, stride=1) |
| 131 | |
| 132 | def forward(self, x): |
| 133 | x = self.conv0(x) |
| 134 | |
| 135 | conv1ori,instance = self.conv1_1(x) |
| 136 | |
| 137 | conv1 = self.lrelu(self.conv1_2(self.lrelu(conv1ori))) |
| 138 | pool1 = self.pool1(conv1) |
| 139 | |