-
Notifications
You must be signed in to change notification settings - Fork 0
/
DenseNet.py
151 lines (132 loc) · 5.18 KB
/
DenseNet.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
141
142
143
144
145
146
147
148
149
150
151
import torch
import torch.nn as nn
from torch.autograd import Variable
from torch.utils.tensorboard import SummaryWriter
class CONV(nn.Module):
def __init__(self,in_channels, out_channels,kernel_size,stride,padding):
super(CONV, self).__init__()
self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, stride, padding,bias=False)
self.bn = nn.BatchNorm2d(out_channels)
self.relu = nn.ReLU()
def forward(self,x):
out = self.conv(x)
out = self.bn(out)
out = self.relu(out)
return out
class Dense_B_layer(nn.Module):
def __init__(self,in_channels, out_channels):
super(Dense_B_layer, self).__init__()
self.bn = nn.BatchNorm2d(in_channels)
self.relu = nn.ReLU()
self.conv1 = CONV(in_channels, out_channels, 1, 1, 0)
self.conv2 = nn.Conv2d(out_channels, out_channels, 3, 1, 1)
def forward(self,x):
out = self.bn(x)
out = self.relu(out)
out = self.conv1(out)
out = self.conv2(out)
out = torch.cat([out, x],dim=1)
return out
class Dense_block(nn.Module):
def __init__(self,in_channels, out_channels, block_scale):
super(Dense_block,self).__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.block_scale = block_scale
self.layers = self.make_layer()
def forward(self,x):
# out = self.layer1(x)
# layers = self.make_layer()
for i in range(len(self.layers)):
if i == 0:
out = self.layers[i](x)
else:
out = self.layers[i](out)
return out[:,0: self.out_channels,:,:]
def make_layer(self):
layer = []
for i in range(self.block_scale):
layer.append(Dense_B_layer((self.in_channels + i*self.out_channels), self.out_channels))
return nn.Sequential(*layer)
class Transition(nn.Module):
def __init__(self,in_channels,out_channels):
super(Transition,self).__init__()
self.bn = nn.BatchNorm2d(in_channels)
self.relu = nn.ReLU()
self.conv = nn.Conv2d(in_channels,out_channels,1,1,0,bias=False)
self.avgpool = nn.AvgPool2d(2,2)
def forward(self,x):
out = self.bn(x)
out = self.relu(out)
out = self.conv(out)
out = self.avgpool(out)
return out
class DenseNet(nn.Module):
def __init__(self, k, theta, block_scale, class_num):
super(DenseNet,self).__init__()
self.conv1 = CONV(3,int(2*k),7,2,3)
self.pool = nn.MaxPool2d(3,2,1)
self.denseblock1 = Dense_block(int(2*k), k, block_scale[0])
self.transition1 = Transition(k, int(theta*k))
self.denseblock2 = Dense_block(int(theta*k), k, block_scale[1])
self.transition2 = Transition(k, int(theta*k))
self.denseblock3 = Dense_block(int(theta*k), k, block_scale[2])
self.transition3 = Transition(k, int(theta*k))
self.denseblock4 = Dense_block(int(theta*k), k, block_scale[3])
self.bn = nn.BatchNorm2d(int(theta*k))
self.relu = nn.ReLU()
self.avgpool = nn.AvgPool2d(7)
self.dropout = nn.Dropout2d(p=0.5)
self.fc = nn.Linear(k,class_num)
for m in self.modules():
if isinstance(m, nn.Conv2d):
nn.init.kaiming_normal_(m.weight)
elif isinstance(m, nn.BatchNorm2d):
nn.init.constant_(m.bias, 0)
nn.init.constant_(m.weight, 1)
elif isinstance(m, nn.Linear):
nn.init.constant_(m.bias, 0)
def forward(self,x):
out = self.conv1(x)
out = self.pool(out)
out = self.denseblock1(out)
out = self.transition1(out)
out = self.denseblock2(out)
out = self.transition2(out)
out = self.denseblock3(out)
out = self.transition3(out)
out = self.denseblock4(out)
out = self.bn(out)
out = self.relu(out)
out = self.avgpool(out)
# out = out.view(out.size(0),-1)
out = torch.flatten(out, 1)
out = self.dropout(out)
out = self.fc(out)
return out
def DenseNet_xxx(model_name, k, theta, class_num):
if model_name == 'DenseNet-121':
return DenseNet(k, theta, [6, 12, 24, 16], class_num)
elif model_name == 'DenseNet-169':
return DenseNet(k, theta, [6, 12, 32, 32], class_num)
elif model_name == 'DenseNet-201':
return DenseNet(k, theta, [6, 12, 48, 32], class_num)
elif model_name == 'DenseNet-264':
return DenseNet(k, theta, [6, 12, 64, 48], class_num)
else:
raise Exception('The {} model does not exsit!'.format(model_name))
def main():
log_path = './summary/'
writer = SummaryWriter(log_dir=log_path, comment='DenseNet')
x = Variable(torch.rand(size=(8,3,224,224)))
model = DenseNet_xxx('DenseNet-121', 12, 1, 1000)
# model_conv = CONV(3,24,7,2,3)
# model_dense_layer = Dense_B_layer(3,12)
# model_dense_block = Dense_block(3,12,6)
# model_trans = Transition(3,12)
# model_dense = DenseNet(12,1,[6, 12, 24, 16],1000)
writer.add_graph(model, (x,), verbose=True)
writer.close()
return
if __name__ == "__main__":
main()