hololens/stable-diffusion-webui-depthmap-script
1
1import torch.nn as nn
2import torch.nn as NN
3
4__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',
5 'resnet152']
6
7
8model_urls = {
9 'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
10 'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',
11 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',
12 'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',
13 'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
14}
15
16
17def conv3x3(in_planes, out_planes, stride=1):
18 """3x3 convolution with padding"""
19 return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
20 padding=1, bias=False)
21
22
23class BasicBlock(nn.Module):
24 expansion = 1
25
26 def __init__(self, inplanes, planes, stride=1, downsample=None):
27 super(BasicBlock, self).__init__()
28 self.conv1 = conv3x3(inplanes, planes, stride)
29 self.bn1 = NN.BatchNorm2d(planes) #NN.BatchNorm2d
30 self.relu = nn.ReLU(inplace=True)
31 self.conv2 = conv3x3(planes, planes)
32 self.bn2 = NN.BatchNorm2d(planes) #NN.BatchNorm2d
33 self.downsample = downsample
34 self.stride = stride
35
36 def forward(self, x):
37 residual = x
38
39 out = self.conv1(x)
40 out = self.bn1(out)
41 out = self.relu(out)
42
43 out = self.conv2(out)
44 out = self.bn2(out)
45
46 if self.downsample is not None:
47 residual = self.downsample(x)
48
49 out += residual
50 out = self.relu(out)
51
52 return out
53
54
55class Bottleneck(nn.Module):
56 expansion = 4
57
58 def __init__(self, inplanes, planes, stride=1, downsample=None):
59 super(Bottleneck, self).__init__()
60 self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False)
61 self.bn1 = NN.BatchNorm2d(planes) #NN.BatchNorm2d
62 self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,
63 padding=1, bias=False)
64 self.bn2 = NN.BatchNorm2d(planes) #NN.BatchNorm2d
65 self.conv3 = nn.Conv2d(planes, planes * self.expansion, kernel_size=1, bias=False)
66 self.bn3 = NN.BatchNorm2d(planes * self.expansion) #NN.BatchNorm2d
67 self.relu = nn.ReLU(inplace=True)
68 self.downsample = downsample
69 self.stride = stride
70
71 def forward(self, x):
72 residual = x
73
74 out = self.conv1(x)
75 out = self.bn1(out)
76 out = self.relu(out)
77
78 out = self.conv2(out)
79 out = self.bn2(out)
80 out = self.relu(out)
81
82 out = self.conv3(out)
83 out = self.bn3(out)
84
85 if self.downsample is not None:
86 residual = self.downsample(x)
87
88 out += residual
89 out = self.relu(out)
90
91 return out
92
93
94class ResNet(nn.Module):
95
96 def __init__(self, block, layers, num_classes=1000):
97 self.inplanes = 64
98 super(ResNet, self).__init__()
99 self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
100 bias=False)
101 self.bn1 = NN.BatchNorm2d(64) #NN.BatchNorm2d
102 self.relu = nn.ReLU(inplace=True)
103 self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
104 self.layer1 = self._make_layer(block, 64, layers[0])
105 self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
106 self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
107 self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
108 #self.avgpool = nn.AvgPool2d(7, stride=1)
109 #self.fc = nn.Linear(512 * block.expansion, num_classes)
110
111 for m in self.modules():
112 if isinstance(m, nn.Conv2d):
113 nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
114 elif isinstance(m, nn.BatchNorm2d):
115 nn.init.constant_(m.weight, 1)
116 nn.init.constant_(m.bias, 0)
117
118 def _make_layer(self, block, planes, blocks, stride=1):
119 downsample = None
120 if stride != 1 or self.inplanes != planes * block.expansion:
121 downsample = nn.Sequential(
122 nn.Conv2d(self.inplanes, planes * block.expansion,
123 kernel_size=1, stride=stride, bias=False),
124 NN.BatchNorm2d(planes * block.expansion), #NN.BatchNorm2d
125 )
126
127 layers = []
128 layers.append(block(self.inplanes, planes, stride, downsample))
129 self.inplanes = planes * block.expansion
130 for i in range(1, blocks):
131 layers.append(block(self.inplanes, planes))
132
133 return nn.Sequential(*layers)
134
135 def forward(self, x):
136 features = []
137
138 x = self.conv1(x)
139 x = self.bn1(x)
140 x = self.relu(x)
141 x = self.maxpool(x)
142
143 x = self.layer1(x)
144 features.append(x)
145 x = self.layer2(x)
146 features.append(x)
147 x = self.layer3(x)
148 features.append(x)
149 x = self.layer4(x)
150 features.append(x)
151
152 return features
153
154
155def resnet18(pretrained=True, **kwargs):
156 """Constructs a ResNet-18 model.
157 Args:
158 pretrained (bool): If True, returns a model pre-trained on ImageNet
159 """
160 model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs)
161 return model
162
163
164def resnet34(pretrained=True, **kwargs):
165 """Constructs a ResNet-34 model.
166 Args:
167 pretrained (bool): If True, returns a model pre-trained on ImageNet
168 """
169 model = ResNet(BasicBlock, [3, 4, 6, 3], **kwargs)
170 return model
171
172
173def resnet50(pretrained=True, **kwargs):
174 """Constructs a ResNet-50 model.
175 Args:
176 pretrained (bool): If True, returns a model pre-trained on ImageNet
177 """
178 model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs)
179
180 return model
181
182
183def resnet101(pretrained=True, **kwargs):
184 """Constructs a ResNet-101 model.
185 Args:
186 pretrained (bool): If True, returns a model pre-trained on ImageNet
187 """
188 model = ResNet(Bottleneck, [3, 4, 23, 3], **kwargs)
189
190 return model
191
192
193def resnet152(pretrained=True, **kwargs):
194 """Constructs a ResNet-152 model.
195 Args:
196 pretrained (bool): If True, returns a model pre-trained on ImageNet
197 """
198 model = ResNet(Bottleneck, [3, 8, 36, 3], **kwargs)
199 return model
200