-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmodel.py
164 lines (135 loc) · 7 KB
/
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
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
import tensorflow as tf
from idcard import gen_id_card
IMAGE_WIDTH = 256
IMAGE_HEIGHT = 32
def batch_norm(x, beta, gamma, phase_train, scope='bn', decay=0.9, eps=1e-5):
"""
数据归一化处理：
https://blog.csdn.net/NNNNNNNNNNNNY/article/details/70331796
http://stackoverflow.com/a/34634291/2267819
:param x:
:param beta:
:param gamma:
:param phase_train:
:param scope:
:param decay:
:param eps:
:return:
"""
with tf.variable_scope(scope):
# beta = tf.get_variable(name='beta', shape=[n_out], initializer=tf.constant_initializer(0.0), trainable=True)
# gamma = tf.get_variable(name='gamma', shape=[n_out], initializer=tf.random_normal_initializer(1.0, stddev), trainable=True)
batch_mean, batch_var = tf.nn.moments(x, [0, 1, 2], name='moments')
ema = tf.train.ExponentialMovingAverage(decay=decay)
def mean_var_with_update():
ema_apply_op = ema.apply([batch_mean, batch_var])
with tf.control_dependencies([ema_apply_op]):
return tf.identity(batch_mean), tf.identity(batch_var)
mean, var = tf.cond(phase_train, mean_var_with_update, lambda: (ema.average(batch_mean), ema.average(batch_var)))
normed = tf.nn.batch_normalization(x, mean, var, beta, gamma, eps)
return normed
# dropout
keep_prob = tf.placeholder(tf.float32)
def cnn(w_alpha=0.01, b_alpha=0.1):
x = tf.placeholder(tf.float32, [-1, IMAGE_HEIGHT, IMAGE_WIDTH, 1])
"""
tf.nn.conv2d 卷积计算
:param x: 输入，shape为[批的大小, 长, 宽, 通道数量]
:param W: 核，shape为[输入长, 输入宽, 输入通道数量, 输出通道数量]
:param stride: 步长，shape为[1,长上移动步长,宽上移动步长,1]
:param padding: 边缘，SAME或VALID，分别指在边缘外部补还是
"""
# 4 conv layer
w_c1 = tf.Variable(w_alpha*tf.random_normal([5, 5, 1, 32]))
b_c1 = tf.Variable(b_alpha*tf.random_normal([32]))
# 256x32x1 =>
conv1 = tf.nn.bias_add(tf.nn.conv2d(x, w_c1, strides=[1, 1, 1, 1], padding='SAME'), b_c1)
conv1 = batch_norm(conv1, tf.constant(0.0, shape=[32]), tf.random_normal(shape=[32], mean=1.0, stddev=0.02), train_phase, scope='bn_1')
conv1 = tf.nn.relu(conv1)
conv1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
conv1 = tf.nn.dropout(conv1, keep_prob)
# 256x32x32
w_c2 = tf.Variable(w_alpha*tf.random_normal([5, 5, 32, 64]))
b_c2 = tf.Variable(b_alpha*tf.random_normal([64]))
conv2 = tf.nn.bias_add(tf.nn.conv2d(conv1, w_c2, strides=[1, 1, 1, 1], padding='SAME'), b_c2)
conv2 = batch_norm(conv2, tf.constant(0.0, shape=[64]), tf.random_normal(shape=[64], mean=1.0, stddev=0.02), train_phase, scope='bn_2')
conv2 = tf.nn.relu(conv2)
conv2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
conv2 = tf.nn.dropout(conv2, keep_prob)
w_c3 = tf.Variable(w_alpha*tf.random_normal([3, 3, 64, 64]))
b_c3 = tf.Variable(b_alpha*tf.random_normal([64]))
conv3 = tf.nn.bias_add(tf.nn.conv2d(conv2, w_c3, strides=[1, 1, 1, 1], padding='SAME'), b_c3)
conv3 = batch_norm(conv3, tf.constant(0.0, shape=[64]), tf.random_normal(shape=[64], mean=1.0, stddev=0.02), train_phase, scope='bn_3')
conv3 = tf.nn.relu(conv3)
conv3 = tf.nn.max_pool(conv3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
conv3 = tf.nn.dropout(conv3, keep_prob)
w_c4 = tf.Variable(w_alpha*tf.random_normal([3, 3, 64, 64]))
b_c4 = tf.Variable(b_alpha*tf.random_normal([64]))
conv4 = tf.nn.bias_add(tf.nn.conv2d(conv3, w_c4, strides=[1, 1, 1, 1], padding='SAME'), b_c4)
conv4 = batch_norm(conv4, tf.constant(0.0, shape=[64]), tf.random_normal(shape=[64], mean=1.0, stddev=0.02), train_phase, scope='bn_4')
conv4 = tf.nn.relu(conv4)
conv4 = tf.nn.max_pool(conv4, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
conv4 = tf.nn.dropout(conv4, keep_prob)
# Fully connected layer
w_d = tf.Variable(w_alpha*tf.random_normal([2*16*64, 1024]))
b_d = tf.Variable(b_alpha*tf.random_normal([1024]))
dense = tf.reshape(conv4, [-1, w_d.get_shape().as_list()[0]])
dense = tf.nn.relu(tf.add(tf.matmul(dense, w_d), b_d))
dense = tf.nn.dropout(dense, keep_prob)
w_out = tf.Variable(w_alpha*tf.random_normal([1024, MAX_CAPTCHA*CHAR_SET_LEN]))
b_out = tf.Variable(b_alpha*tf.random_normal([MAX_CAPTCHA*CHAR_SET_LEN]))
out = tf.add(tf.matmul(dense, w_out), b_out)
# out = tf.nn.softmax(out)
return out
# 生成一个训练batch
def get_next_batch(batch_size=128):
obj = gen_id_card()
batch_x = np.zeros([batch_size, IMAGE_HEIGHT*IMAGE_WIDTH])
batch_y = np.zeros([batch_size, MAX_CAPTCHA*CHAR_SET_LEN])
for i in range(batch_size):
image, text, vec = obj.gen_image()
batch_x[i,:] = image.reshape((IMAGE_HEIGHT*IMAGE_WIDTH))
batch_y[i,:] = vec
return batch_x, batch_y
####################################################################
X = tf.placeholder(tf.float32, [None, IMAGE_HEIGHT*IMAGE_WIDTH])
Y = tf.placeholder(tf.float32, [None, MAX_CAPTCHA*CHAR_SET_LEN])
keep_prob = tf.placeholder(tf.float32) # dropout
# 是否在训练阶段
train_phase = tf.placeholder(tf.bool)
# 训练
def train_cnn():
output = cnn()
# loss
#loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=output, labels=Y))
loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=output, labels=Y))
# 最后一层用来分类的softmax和sigmoid有什么不同？
# optimizer 为了加快训练 learning_rate应该开始大，然后慢慢衰
optimizer = tf.train.AdamOptimizer(learning_rate=0.002).minimize(loss)
predict = tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN])
max_idx_p = tf.argmax(predict, 2)
max_idx_l = tf.argmax(tf.reshape(Y, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
correct_pred = tf.equal(max_idx_p, max_idx_l)
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
step = 0
while True:
batch_x, batch_y = get_next_batch(64)
_, loss_ = sess.run([optimizer, loss], feed_dict={X: batch_x,
Y: batch_y, keep_prob: 0.75, train_phase: True})
print(step, loss_)
# 每100 step计算一次准确率
if step % 100 == 0 and step != 0:
batch_x_test, batch_y_test = get_next_batch(100)
acc = sess.run(accuracy, feed_dict={X: batch_x_test,
Y: batch_y_test, keep_prob: 1., train_phase: False})
print("第%s步，训练准确率为：%s" % (step, acc))
# 如果准确率大80%,保存模型,完成训练
if acc > 0.8:
saver.save(sess, "crack_capcha.model", global_step=step)
break
step += 1
if __name__ == '__main__':
train_cnn()