This repository has been archived by the owner on Apr 4, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathvgg.py
155 lines (120 loc) · 6.27 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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
# -*- coding: utf-8 -*-
import scipy.io
import tensorflow as tf
# for mask
vgg_layers = (
'conv1_1', 'relu1_1', 'conv1_2', 'relu1_2', 'pool1',
'conv2_1', 'relu2_1', 'conv2_2', 'relu2_2', 'pool2',
'conv3_1', 'relu3_1', 'conv3_2', 'relu3_2', 'conv3_3',
'relu3_3', 'conv3_4', 'relu3_4', 'pool3',
'conv4_1', 'relu4_1', 'conv4_2', 'relu4_2', 'conv4_3',
'relu4_3', 'conv4_4', 'relu4_4', 'pool4',
'conv5_1', 'relu5_1', 'conv5_2', 'relu5_2', 'conv5_3',
'relu5_3', 'conv5_4', 'relu5_4'
)
'''
Usage:
we only use vgg_structure().build(input_image)
Parameters: input_image: nparray
e.g.
After:
structure = vgg_structure()
structure.build(input_image)
structure.conv1_1 will give the output (a tensorflow variable) of layer "conv1_1"
'''
def file_vgg_read(pretrianed_vgg):
mat = scipy.io.loadmat(pretrianed_vgg)
layers_info = mat["layers"]
# normalization_info = mat["normalization"]
return layers_info #, normalization_info
def conv2d_relu(layers_info, prev_layer, layer_id):
W = layers_info[0][layer_id][0][0][2][0][0]
b = layers_info[0][layer_id][0][0][2][0][1]
b = b.reshape(b.size)
#W = tf.Variable(W, trainable = False)
#b = tf.Variable(b, trainable = False)
W = tf.constant(W)
b = tf.constant(b)
conv2d = tf.nn.conv2d(input=prev_layer, \
filter=W, \
strides=[1, 1, 1, 1], \
padding="SAME")
output = tf.nn.relu(conv2d + b)
return output
def pool(prev_layer):
output = tf.nn.avg_pool(value=prev_layer, \
ksize=[1, 2, 2, 1], \
strides=[1, 2, 2, 1], \
padding="SAME")
# if use max pooling, use this output:****************************************************
'''
output = tf.nn.max_pool(value=prev_layer, \
ksize=[1, 2, 2, 1], \
strides=[1, 2, 2, 1], \
padding="SAME")
'''
return output
class vgg_structure():
def build(self, input_image, layers_info):
#layers_info = file_vgg_read("imagenet-vgg-verydeep-19.mat")
# build the vgg training net
input_layer = input_image
setattr(self, "conv1_1" , conv2d_relu(layers_info, prev_layer = input_layer, layer_id = 0) )
setattr(self, "conv1_2" , conv2d_relu(layers_info, prev_layer = self.conv1_1, layer_id = 2) )
setattr(self, "pool_1" , pool(prev_layer = self.conv1_2) )
setattr(self, "conv2_1" , conv2d_relu(layers_info, prev_layer = self.pool_1, layer_id = 5) )
setattr(self, "conv2_2" , conv2d_relu(layers_info, prev_layer = self.conv2_1, layer_id = 7) )
setattr(self, "pool_2" , pool(prev_layer = self.conv2_2) )
setattr(self, "conv3_1" , conv2d_relu(layers_info, prev_layer = self.pool_2, layer_id = 10) )
setattr(self, "conv3_2" , conv2d_relu(layers_info, prev_layer = self.conv3_1, layer_id = 12) )
setattr(self, "conv3_3" , conv2d_relu(layers_info, prev_layer = self.conv3_2, layer_id = 14) )
setattr(self, "conv3_4" , conv2d_relu(layers_info, prev_layer = self.conv3_3, layer_id = 16) )
setattr(self, "pool_3" , pool(prev_layer = self.conv3_4) )
setattr(self, "conv4_1" , conv2d_relu(layers_info, prev_layer = self.pool_3, layer_id = 19) )
setattr(self, "conv4_2" , conv2d_relu(layers_info, prev_layer = self.conv4_1, layer_id = 21) )
setattr(self, "conv4_3" , conv2d_relu(layers_info, prev_layer = self.conv4_2, layer_id = 23) )
setattr(self, "conv4_4" , conv2d_relu(layers_info, prev_layer = self.conv4_3, layer_id = 25) )
setattr(self, "pool_4" , pool(prev_layer = self.conv4_4) )
setattr(self, "conv5_1" , conv2d_relu(layers_info, prev_layer = self.pool_4, layer_id = 28) )
setattr(self, "conv5_2" , conv2d_relu(layers_info, prev_layer = self.conv5_1, layer_id = 30) )
setattr(self, "conv5_3" , conv2d_relu(layers_info, prev_layer = self.conv5_2, layer_id = 32) )
setattr(self, "conv5_4" , conv2d_relu(layers_info, prev_layer = self.conv5_3, layer_id = 34) )
setattr(self, "pool_5" , pool(prev_layer = self.conv5_4) )
def buildmask(self, input_mask, mask_downsample_type):
net = {}
current = input_mask
# soft
if mask_downsample_type == 'simple':
for name in vgg_layers:
layer_kind = name[:4]
if layer_kind == 'pool':
current = tf.nn.avg_pool(current, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')
setattr(self, name, current)
# hard
elif mask_downsample_type == 'all':
for name in vgg_layers:
layer_kind = name[:4]
if layer_kind == 'conv':
current = tf.nn.max_pool(current, ksize=[1,3,3,1], strides=[1,1,1,1], padding='SAME')
elif layer_kind == 'pool':
current = tf.nn.max_pool(current, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')
setattr(self, name, current)
# hard, keep the padding boundary unchanged
elif mask_downsample_type == 'inside':
current = 1 - current
for name in vgg_layers:
layer_kind = name[:4]
if layer_kind == 'conv':
current = tf.nn.max_pool(current, ksize=[1,3,3,1], strides=[1,1,1,1], padding='SAME')
elif layer_kind == 'pool':
current = tf.nn.max_pool(current, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')
setattr(self, name, current)
# soft
elif mask_downsample_type == 'mean':
for name in vgg_layers:
layer_kind = name[:4]
if layer_kind == 'conv':
current = tf.nn.avg_pool(current, ksize=[1,3,3,1], strides=[1,1,1,1], padding='SAME')
elif layer_kind == 'pool':
current = tf.nn.avg_pool(current, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')
setattr(self, name, current)