-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathVGG.py
136 lines (97 loc) · 7.52 KB
/
VGG.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
import tensorflow as tf
import tools
#%%
def VGG16(x, n_classes, is_pretrain=True): #is_pretrain为False则固定住参数不改变
x = tools.conv('conv1_1', x, out_channels=64, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv1_2', x, out_channels=64, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.pool('pool1', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv2_1', x, out_channels=128, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv2_2', x, out_channels=128, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.pool('pool2', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv3_1', x, out_channels=256, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv3_2', x, out_channels=256, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv3_3', x, out_channels=256, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.pool('pool3', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv4_1', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv4_2', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv4_3', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.pool('pool3', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv5_1', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv5_2', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv5_3', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.pool('pool3', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.FC_layer('fc6', x, out_nodes=1024)
x = tools.batch_norm(x) # or dropout
x = tools.FC_layer('fc7', x, out_nodes=1024)
x = tools.batch_norm(x) # or dropout
x = tools.FC_layer('fc8', x, out_nodes=n_classes)
return x
#%% TO get better tensorboard figures!
def VGG16N(x, n_classes, is_pretrain=True): # True则可在训练时改变
with tf.name_scope('VGG16'):
x = tools.conv('conv1_1', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv1_2', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
with tf.name_scope('pool1'):
x = tools.pool('pool1', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv2_1', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv2_2', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
with tf.name_scope('pool2'):
x = tools.pool('pool2', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv3_1', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv3_2', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv3_3', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
with tf.name_scope('pool3'):
x = tools.pool('pool3', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv4_1', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv4_2', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv4_3', x, out_channels=32, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
with tf.name_scope('pool4'):
x = tools.pool('pool4', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.conv('conv5_1', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv5_2', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
x = tools.conv('conv5_3', x, out_channels=512, kernel_size=[3,3,3], stride=[1,1,1,1,1], is_pretrain=is_pretrain)
with tf.name_scope('pool5'):
x = tools.pool('pool5', x, kernel=[1,8,8,8,1], stride=[1,8,8,8,1], is_max_pool=True)
x = tools.FC_layer('fc6', x, out_nodes=256)
with tf.name_scope('batch_norm1'):
x = tools.batch_norm(x)
x = tools.FC_layer('fc7', x, out_nodes=256)
with tf.name_scope('batch_norm2'):
x = tools.batch_norm(x)
x = tools.FC_layer('fc8', x, out_nodes=n_classes)
return x
# 改动第一层以便迁移学习
def VGG16T(x, n_classes, is_pretrain=True): # True则可在训练时改变
with tf.name_scope('VGG16'):
x = tools.conv2D('conv1_1', x, 64, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=is_pretrain)
x = tools.conv2D('conv1_2', x, 64, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
with tf.name_scope('pool1'):
x = tools.pool2D('pool1', x, kernel=[1, 2, 2, 1], stride=[1, 2, 2, 1], is_max_pool=True)
x = tools.conv2D('conv2_1', x, 128, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
x = tools.conv2D('conv2_2', x, 128, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
with tf.name_scope('pool2'):
x = tools.pool2D('pool2', x, kernel=[1, 2, 2, 1], stride=[1, 2, 2, 1], is_max_pool=True)
x = tools.conv2D('conv3_1', x, 256, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
x = tools.conv2D('conv3_2', x, 256, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
x = tools.conv2D('conv3_3', x, 256, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
with tf.name_scope('pool3'):
x = tools.pool2D('pool3', x, kernel=[1, 2, 2, 1], stride=[1, 2, 2, 1], is_max_pool=True)
x = tools.conv2D('conv4_1', x, 512, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
x = tools.conv2D('conv4_2', x, 512, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
x = tools.conv2D('conv4_3', x, 512, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
with tf.name_scope('pool4'):
x = tools.pool2D('pool4', x, kernel=[1, 2, 2, 1], stride=[1, 2, 2, 1], is_max_pool=True)
x = tools.conv2D('conv5_1', x, 512, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
x = tools.conv2D('conv5_2', x, 512, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
x = tools.conv2D('conv5_3', x, 512, kernel_size=[3, 3], stride=[1, 1, 1, 1], is_pretrain=False)
with tf.name_scope('pool5'):
x = tools.pool2D('pool5', x, kernel=[1, 2, 2, 1], stride=[1, 2, 2, 1], is_max_pool=True)
x = tools.FC_layer('fc6', x, out_nodes=4096)
with tf.name_scope('batch_norm1'):
x = tools.batch_norm(x)
x = tools.FC_layer('fc7', x, out_nodes=4096)
with tf.name_scope('batch_norm2'):
x = tools.batch_norm(x)
x = tools.FC_layer('fc8', x, out_nodes=n_classes)
return x
#%%