forked from chenxuluo/GST-video
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathSTM.py
179 lines (148 loc) · 7.73 KB
/
STM.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
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
import torch
import torch.nn as nn
import math
import torch.utils.model_zoo as model_zoo
__all__ = ['ResNet', 'resnet50']
model_urls = {
'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',
'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',
'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',
'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
}
class CMM_Channel_wise2D(nn.Module):
def __init__(self,planes,stride):
super(CMM_Channel_wise2D, self).__init__()
self.Channel_wise_2d = nn.Conv3d(in_channels=planes,out_channels=planes,
kernel_size=(1,3,3),groups=planes,
padding=(0,1,1),stride=(1,stride,stride),
bias=False)
class Bottleneck(nn.Module):
expansion = 4
def __init__(self, inplanes, planes, stride=1, downsample=None):
super(Bottleneck, self).__init__()
self.conv1 = nn.Conv3d(inplanes, planes, kernel_size=1, bias=False) # 定义一个逐点卷积进行通道融合
self.bn1 = nn.BatchNorm3d(planes)
#Input_channel:planes ==> output_channels:planes
#============================================================================
"""定义CSTM模块"""
# 定义2D的空间卷积
#定义逐通道的1D时域卷积，按照文章中所述进行初始化，前1/4为[1,0,0],后1/4为[0,0,1],另一半为[0,1,0]
self.channel_wise_1D = nn.Conv3d(in_channels=planes,out_channels=planes,
kernel_size=(3,1,1),groups=planes,
padding=(1,0,0),stride=(1,1,1),
bias=False)
#注意padding和stride的设置使得featuremap的尺寸受到变量stride的控制。
# 定义2D空间卷积,用Imagenet进行初始化
self.conv2 = nn.Conv3d(in_channels=planes,out_channels=planes,
kernel_size=(1,3,3),stride=(1,stride,stride),
padding=(0,1,1),bias=False)
"""定义CMM模块"""
self.CMM_2d = nn.Conv3d(in_channels=planes,out_channels=planes/16,
kernel_size=(1,1,1),stride=(1,1,1),padding=(0,0,0),
bias=False)
self.channel_2D_1 = self.CMM_Channel_wise2D(planes/16,stride)
#============================================================================
self.bn2 = nn.BatchNorm3d(planes)
self.conv3 = nn.Conv3d(planes, planes * self.expansion, kernel_size=1, bias=False) # 再一次进行3逐点卷积扩充通道
self.bn3 = nn.BatchNorm3d(planes * self.expansion) # 在一次进行3DBN
self.relu = nn.ReLU(inplace=True)
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x #保留住残差
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out) # out ==> [batch_size,C,T,W,H]
out = self.channel_wise_1D(out)
out = self.conv2(out)
out = self.bn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.bn3(out)
if self.downsample is not None:
#降采样的目的是是为下一步将residual与残差块卷积计算出来的特征图进行相加做准备
#因为经过残差卷积后通道数量已经比residual多了，因此需要进行处理。
residual = self.downsample(residual)
out += residual # 加上残差
out = self.relu(out) # 再进行激活一次
return out
def CMM_Channel_wise2D(self,planes,stride):
return nn.Sequential(nn.Conv3d(in_channels=planes,out_channels=planes,
kernel_size=(1,3,3),stride=(1,stride,stride),
padding=(0,1,1),groups=planes))
class ResNet(nn.Module):
def __init__(self, block, layers, num_classes=174):
self.inplanes = 64
super(ResNet, self).__init__()
self.conv1 = nn.Conv3d(3, 64, kernel_size=(1, 7, 7), stride=(1, 2, 2), padding=(0, 3, 3),
bias=False)
self.bn1 = nn.BatchNorm3d(64)
self.relu = nn.ReLU(inplace=True)# 定义Relu激活,输出featuremap的特征图的通道数为64
self.maxpool = nn.MaxPool3d(kernel_size=(1, 3, 3), stride=(1, 2, 2), padding=(0, 1, 1))
#此处self._make_layer函数的参数（64,128,256,512）
# 表示的是bottleneck连接中第二个卷积层（3X3）卷积的输入和输出通道数
#且bottleneck的第三个卷积（1X1的通道卷积的输出通道数量为，expansionX(64,128,256,512)
self.layer1 = self._make_layer(block, 64, layers[0])
self.layer2 = self._make_layer(block, 128, layers[1],stride=2)
self.layer3 = self._make_layer(block, 256, layers[2],stride=2)
self.layer4 = self._make_layer(block, 512, layers[3],stride=2)
self.avgpool = nn.AvgPool2d(7, stride=1)
self.new_fc = nn.Linear(512 * block.expansion, num_classes)
for m in self.modules():
if isinstance(m, nn.Conv3d):
nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') #对所有的卷积层使用凯明初始化
elif isinstance(m, nn.BatchNorm3d):#所有的BN层采用gamma=1,beta=0
nn.init.constant_(m.weight, 1)
nn.init.constant_(m.bias, 0)
def _make_layer(self, block, planes, blocks, stride=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv3d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=(1, stride, stride), bias=False),
nn.BatchNorm3d(planes * block.expansion),
)
layers = []
# 只有每一组卷积的第一个bottneck连接需要用到downsample
layers.append(block(self.inplanes, planes, stride, downsample))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def forward(self, x):
x = self.conv1(x)
x = self.bn1(x)
x = self.relu(x)
x = self.maxpool(x) #此时featuremap为64
print("maxpool.shape",x.size())
x = self.layer1(x)
print("layer1_out.size:",x.size())
x = self.layer2(x)
print("layer2_out.size:",x.size())
x = self.layer3(x)
print("layer3_out.size:",x.size())
x = self.layer4(x)
print("layer4_out.size:",x.size())
x = x.transpose(1, 2).contiguous() #从此处开始
x = x.view((-1,) + x.size()[2:])
x = self.avgpool(x)
x = x.view(x.size(0), -1)
x = self.new_fc(x)
print("out.size():",x.size())
return x
def resnet50(**kwargs):
"""Constructs a ResNet-50 based model.
"""
model = ResNet(Bottleneck, [3, 4, 6, 3],**kwargs)
checkpoint = model_zoo.load_url(model_urls['resnet50'],model_dir='./pretrained/') # 加载Imagenet预训练的模型参数
layer_name = list(checkpoint.keys()) # checkpoint是一个字典{'layer_name','权重tensor'}
for ln in layer_name:
if 'conv' in ln or 'downsample.0.weight' in ln:
checkpoint[ln] = checkpoint[ln].unsqueeze(2) # 扩充维度，很重要！
model.load_state_dict(checkpoint, strict=False)
return model
if __name__ == '__main__':
model = resnet50()
Input = torch.randn([1,3,8,224,224]) #N,C,T,W,H
out = model(Input)