-
Notifications
You must be signed in to change notification settings - Fork 17
/
Net.py
135 lines (105 loc) · 6.08 KB
/
Net.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
# author is He Zhao
# The time to create is 8:47 PM, 28/11/16
import tensorflow as tf
import tensorflow.contrib.slim as slim
from Opts import lrelu, resUnit
from dataBlocks import DataBlocks
import pickle
import scipy.io as sio
import numpy as np
initializer = tf.truncated_normal_initializer(stddev=0.02)
bias_initializer = tf.constant_initializer(0.0)
def build_data(batchsize, dataset_name='DRIVE'):
#readpath = open('img.pkl', 'rb')
#datapaths = pickle.load(readpath)
datapaths = [['data/{}_train_image.npy'.format(dataset_name),
'data/{}_train_gt.npy'.format(dataset_name),
'data/{}_train_mask.npy'.format(dataset_name)]]
db = DataBlocks(data_paths=datapaths, train_valid_ratio=[39, 0], batchsize=batchsize, allow_preload=False)
return db
def matTonpy_35():
img = sio.loadmat('test_1To4.mat')['imgAllTest']
gt = sio.loadmat('test_1To4.mat')['gtAllTest']
mask = sio.loadmat('test_1To4.mat')['maskAllTest']
return img, gt, mask
def discriminator(image, reuse=False):
n=32
bn = slim.batch_norm
with tf.variable_scope("discriminator"):
# original
dis1 = slim.convolution2d(image, n, [4, 4], 2, activation_fn=lrelu,
reuse=reuse, scope='d_conv1', weights_initializer=initializer) # 256 256 64
dis2 = slim.convolution2d(dis1, 2*n, [4, 4], 2, normalizer_fn=bn, activation_fn=lrelu,
reuse=reuse, scope='d_conv2', weights_initializer=initializer) # 128 128 64
dis3 = slim.convolution2d(dis2, 4*n, [4, 4], 2, normalizer_fn=bn, activation_fn=lrelu,
reuse=reuse, scope='d_conv3', weights_initializer=initializer) # 64 64 128
dis4 = slim.convolution2d(dis3, 8*n, [4, 4], 2, normalizer_fn=bn, activation_fn=lrelu,
reuse=reuse, scope='d_conv4', weights_initializer=initializer) # 32 32 256
dis5 = slim.convolution2d(dis4, 16*n, [4, 4], 2, normalizer_fn=bn, activation_fn=lrelu,
reuse=reuse, scope='d_conv5', weights_initializer=initializer) # 16 16 512
d_out_logits = slim.fully_connected(slim.flatten(dis5), 1, activation_fn=None, reuse=reuse, scope='d_out',
weights_initializer=initializer)
d_out = tf.nn.sigmoid(d_out_logits)
return d_out, d_out_logits
def generator(image, act_input, z):
n = 64
with tf.variable_scope("generator"):
# original
e1 = slim.conv2d(image, n, [4, 4], 2, activation_fn=lrelu, scope='g_e1_conv',
weights_initializer=initializer)
# 256
e2 = slim.conv2d(tf.concat([lrelu(e1), act_input[256]], 3), 2 * n, [4, 4], 2, normalizer_fn=slim.batch_norm, activation_fn=None, scope='g_e2_conv',
weights_initializer=initializer)
# 128
e3 = slim.conv2d(lrelu(e2), 4 * n, [4, 4], 2, normalizer_fn=slim.batch_norm, activation_fn=None, scope='g_e3_conv',
weights_initializer=initializer)
# 64
e4 = slim.conv2d(tf.concat([lrelu(e3), act_input[64]], 3), 8 * n, [4, 4], 2, normalizer_fn=slim.batch_norm, activation_fn=None, scope='g_e4_conv',
weights_initializer=initializer)
# 32
e5 = slim.conv2d(lrelu(e4), 8*n, [4, 4], 2, normalizer_fn=slim.batch_norm, activation_fn=None, scope='g_e5_conv',
weights_initializer=initializer)
# 16
e6 = slim.conv2d(lrelu(e5), 8*n, [4, 4], 2, normalizer_fn=slim.batch_norm, activation_fn=None, scope='g_e6_conv',
weights_initializer=initializer)
# 8
zP = slim.fully_connected(z, 4 * 4 * n, normalizer_fn=None, activation_fn=lrelu, scope='g_project',
weights_initializer=initializer)
zCon = tf.reshape(zP, [-1, 4, 4, n])
gen1 = slim.conv2d(tf.image.resize_nearest_neighbor(lrelu(zCon), [8, 8]),
2 * n, [3, 3], 1, padding='SAME', normalizer_fn=slim.batch_norm, activation_fn=None,
scope='g_dconv1', weights_initializer=initializer)
# 8
gen1 = tf.concat([gen1, e6], 3)
gen2 = slim.conv2d(tf.image.resize_nearest_neighbor(lrelu(gen1), [16, 16]),
4 * n, [3, 3], 1, normalizer_fn=slim.batch_norm, activation_fn=None,
scope='g_dconv2', weights_initializer=initializer)
# 16
gen2 = tf.concat([gen2, e5], 3)
gen3 = slim.conv2d(tf.image.resize_nearest_neighbor(lrelu(gen2), [32, 32]),
8 * n, [3, 3], 1, normalizer_fn=slim.batch_norm, activation_fn=None,
scope='g_dconv3', weights_initializer=initializer)
gen3 = tf.concat([gen3, e4], 3)
# 32
gen6 = slim.conv2d(tf.image.resize_nearest_neighbor(tf.nn.relu(gen3), [64, 64]),
4 * n, [3, 3], 1, normalizer_fn=slim.batch_norm, activation_fn=None,
scope='g_dconv6', weights_initializer=initializer)
gen6 = tf.concat([gen6, e3], 3)
# 64
gen7 = slim.conv2d(tf.image.resize_nearest_neighbor(tf.nn.relu(gen6), [128, 128]),
2 * n, [3, 3], 1, normalizer_fn=slim.batch_norm, activation_fn=None,
scope='g_dconv7', weights_initializer=initializer)
gen7 = tf.concat([gen7, e2], 3)
# 128
gen8 = slim.conv2d(tf.image.resize_nearest_neighbor(tf.nn.relu(gen7), [256, 256]),
n, [3, 3], 1, normalizer_fn=slim.batch_norm, activation_fn=None,
scope='g_dconv8', weights_initializer=initializer)
# 256
# gen8 = tf.nn.dropout(gen8, 0.5)
gen8 = tf.concat([gen8, e1], 3)
gen8 = tf.nn.relu(gen8)
# 256
gen_out = slim.conv2d(tf.image.resize_nearest_neighbor(gen8, [512, 512]),
3, [3, 3], 1, activation_fn=tf.nn.tanh, scope='g_out',
weights_initializer=initializer)
return gen_out