-
Notifications
You must be signed in to change notification settings - Fork 1
/
model.py
142 lines (107 loc) · 4.12 KB
/
model.py
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
import torch
import torchvision
import torch.nn as nn
import matplotlib.pyplot as plt
import torch.utils.data as Data
class LowLevelNet(nn.Module):
def __init__(self):
super(LowLevelNet, self).__init__()
self.conv1=nn.Conv2d(in_channels=1,out_channels=64,kernel_size=3,stride=2,padding=1)
self.conv2=nn.Conv2d(64,128,3,1,1)
self.conv3=nn.Conv2d(128,128,3,2,1)
self.conv4=nn.Conv2d(128,256,3,1,1)
self.conv5=nn.Conv2d(256,256,3,2,1)
self.conv6=nn.Conv2d(256,512,3,1,1)
def forward(self,x):
out = nn.ReLU()(self.conv1(x))
out = nn.ReLU()(self.conv2(out))
out = nn.ReLU()(self.conv3(out))
out = nn.ReLU()(self.conv4(out))
out = nn.ReLU()(self.conv5(out))
out = nn.ReLU()(self.conv6(out))
return out
class MidLevelNet(nn.Module):
def __init__(self):
super(MidLevelNet, self).__init__()
self.conv1=nn.Conv2d(512,512,3,1,1)
self.conv2=nn.Conv2d(512,256,3,1,1)
def forward(self,x):
out = nn.ReLU()(self.conv1(x))
out = nn.ReLU()(self.conv2(out))
return out
class GlobalNet(nn.Module):
def __init__(self):
super(GlobalNet, self).__init__()
self.conv1=nn.Conv2d(512,512,3,2,1)
self.conv2=nn.Conv2d(512,512,3,1,1)
self.conv3=nn.Conv2d(512,512,3,2,1)
self.conv4=nn.Conv2d(512,512,3,1,1)
self.fc1=nn.Linear(7*7*512,1024)#如果按照论文的数据集7*7*
self.fc2=nn.Linear(1024,512)
self.fc3=nn.Linear(512,256)
def forward(self,x):
out = nn.ReLU()(self.conv1(x))
out = nn.ReLU()(self.conv2(out))
out = nn.ReLU()(self.conv3(out))
out = nn.ReLU()(self.conv4(out))
out = out.view(-1, 7 * 7 * 512) # 如果按照论文的数据集7*7*
out = nn.ReLU()(self.fc1(out))
out = nn.ReLU()(self.fc2(out))
classIn = out
out = nn.ReLU()(self.fc3(out))
fusionIn = out
return fusionIn,classIn
class ClassNet(nn.Module):
def __init__(self,numClasses):
super(ClassNet, self).__init__()
self.fc1=nn.Linear(512,256)
self.fc2=nn.Linear(256,numClasses)
def forward(self,x):
out = nn.ReLU()(self.fc1(x))
out=self.fc2(out)
return out
class ColorizeNet(nn.Module):
def __init__(self):
super(ColorizeNet, self).__init__()
self.conv1=nn.Conv2d(512,128,3,1,1)
self.conv2=nn.Conv2d(128,64,3,1,1)
self.conv3=nn.Conv2d(64,64,3,1,1)
self.conv4=nn.Conv2d(64,32,3,1,1)
self.conv5=nn.Conv2d(32,2,3,1,1)
def forward(self,x):
out = nn.ReLU()(self.conv1(x))
out=nn.functional.interpolate(out,scale_factor=2)
out = nn.ReLU()(self.conv2(out))
out = nn.ReLU()(self.conv3(out))
out=nn.functional.interpolate(out,scale_factor=2)
out = nn.ReLU()(self.conv4(out))
out = nn.Sigmoid()(self.conv5(out))
out=nn.functional.interpolate(out,scale_factor=2) #L*a*b
return out
class Net(nn.Module):
def __init__(self,num_classes):
super(Net,self).__init__()
self.lowLevel=LowLevelNet()
self.midLevel=MidLevelNet()
self.globalLevel=GlobalNet()
#融合
self.classNet=ClassNet(num_classes)
self.colorize=ColorizeNet()
def fusionLayer(self, midOut, globalOut):
#midOut (batchsize,256,h,w)
#globalOut (batchsize,256,1,1)
batch_size, num = globalOut.shape
globalOut = globalOut.reshape((batch_size, num, 1, 1))
tmp = torch.repeat_interleave(globalOut, repeats=midOut.shape[2], dim=2)
tmp = torch.repeat_interleave(tmp, repeats=midOut.shape[3], dim=3)
fuseOut = torch.cat([tmp, midOut], dim=1)
return nn.Sigmoid()(fuseOut)
def forward(self,x):
lowNetOut=self.lowLevel(x)
midNetOut=self.midLevel(lowNetOut)
fusionIn,classIn=self.globalLevel(lowNetOut)
classOut=self.classNet(classIn)
#下面是融合层 之后进入着色层，输出out
fuseOut=self.fusionLayer(midNetOut, fusionIn)
out=self.colorize(fuseOut)
return out, classOut