-
Notifications
You must be signed in to change notification settings - Fork 0
/
C3D_model.py
122 lines (91 loc) · 4.61 KB
/
C3D_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
import torch
import torch.nn as nn
class C3D(nn.Module):
def __init__(self, num_classes,prtrained=True):
super(C3D, self).__init__()
self.pretrained = prtrained
self.conv1 = nn.Conv3d(3, 64, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第一个卷积层
self.pool1 = nn.MaxPool3d(kernel_size=(1, 2, 2), stride=(1, 2, 2)) # 第一个池化层
self.conv2 = nn.Conv3d(64, 128, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第二个卷积层
self.pool2 = nn.MaxPool3d(kernel_size=(2, 2, 2), stride=(2, 2, 2)) # 第二个池化层
self.conv3a = nn.Conv3d(128, 256, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第三个卷积层
self.conv3b = nn.Conv3d(256, 256, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第三个卷积层
self.pool3 = nn.MaxPool3d(kernel_size=(2, 2, 2), stride=(2, 2, 2)) # 第三个池化层
self.conv4a = nn.Conv3d(256, 512, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第四个卷积层
self.conv4b = nn.Conv3d(512, 512, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第四个卷积层
self.pool4 = nn.MaxPool3d(kernel_size=(2, 2, 2), stride=(2, 2, 2)) # 第四个池化层
self.conv5a = nn.Conv3d(512, 512, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第五个卷积层
self.conv5b = nn.Conv3d(512, 512, kernel_size=(3, 3, 3), padding=(1, 1, 1)) # 第五个卷积层
self.pool5 = nn.MaxPool3d(kernel_size=(2, 2, 2), stride=(2, 2, 2), padding=(0, 1, 1)) # 第五个池化层
self.fc6 = nn.Linear(8192, 4096)
self.fc7 = nn.Linear(4096, 4096)
self.fc8 = nn.Linear(4096, num_classes)
self.relu = nn.ReLU()
self.dropout = nn.Dropout(p=0.5) # 随机失活，也就是使得有些参数不更新
self.__init_weight()
if self.pretrained:
self.__load__pretrained_weights()
def forward(self, x):
# 数据格式为 channels * flash * h * w
x = self.relu(self.conv1(x))
x = self.pool1(x)
x = self.relu(self.conv2(x))
x = self.pool2(x)
x = self.relu(self.conv3a(x))
x = self.relu(self.conv3b(x))
x = self.pool3(x)
x = self.relu(self.conv4a(x))
x = self.relu(self.conv4b(x))
x = self.pool4(x)
x = self.relu(self.conv5a(x))
x = self.relu(self.conv5b(x))
x = self.pool5(x)
x = x.reshape(-1, 8192) # 原来x是一个四维的，所以需要我们把他转换成一维的，结果为8192，1
x = self.relu(self.fc6(x))
x = self.dropout(x)
x = self.relu(self.fc7(x))
x = self.dropout(x)
x = self.fc8(x)
return x
def __init_weight(self):
for m in self.modules():
if isinstance(m, nn.Conv3d): # 检测是否都是3D的，如果是则手动初始化防止初始化过于随机
torch.nn.init.kaiming_normal(m.weight)
def __load__pretrained_weights(self):
corresp_name = {
'features.0.weight': 'conv1.weight',
'features.0.bias': 'conv1.bias',
'features.3.weight': 'conv2.weight',
'features.3.bias': 'conv2.bias',
'features.6.weight': 'conv3a.weight',
'features.6.bias': 'conv3a.bias',
'features.8.weight': 'conv3b.weight',
'features.8.bias': 'conv3b.bias',
'features.11.weight': 'conv4a.weight',
'features.11.bias': 'conv4a.bias',
'features.13.weight': 'conv4b.weight',
'features.13.bias': 'conv4b.bias',
'features.16.weight': 'conv5a.weight',
'features.16.bias': 'conv5a.bias',
'features.18.weight': 'conv5b.weight',
'features.18.bias': 'conv5b.bias',
'classifier.0.weight': 'fc6.weight',
'classifier.0.bias': 'fc6.bias',
'classifier.3.weight': 'fc7.weight',
'classifier.3.bias': 'fc7.bias',
}
p_dict = torch.load('ucf101-caffe.pth')
s_dict = self.state_dict() # 这段是显示自己的初始化参数，使用的是kaimin初始法
# print(s_dict['conv1.weight'])
# 循环替换模型参数
for name in p_dict:
if name not in corresp_name:
continue
s_dict[corresp_name[name]] = p_dict[name]
self.load_state_dict(s_dict)
if __name__ == '__main__':
from torchsummary import summary
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# inputs = torch.rand(1, 3, 16, 112, 112) # batch，channel， flash， w， h
net = C3D(num_classes=101).to(device)
# print(summary(net, (3, 16, 112, 112)))