-
Notifications
You must be signed in to change notification settings - Fork 1
/
pre.py
executable file
·204 lines (183 loc) · 6.95 KB
/
pre.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
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
#coding=utf-8
import cv2
import os
import shutil
import numpy as np
import pandas as pd
import lmdb
import caffe
def binary_img(img):
"""
二值化
:return: 二值化之后的图像
"""
retval, t = cv2.threshold(img, 125, 1, cv2.THRESH_BINARY)
h_sum = t.sum(axis=0)
v_sum = t.sum(axis=1)
x1, x2 = (v_sum > 1).nonzero()[0][0], (v_sum > 1).nonzero()[0][-1]
y1, y2 = (h_sum > 5).nonzero()[0][0], (h_sum > 1).nonzero()[0][-1]
im = img[x1:x2, y1:y2]
return im
def ding_ge(binary_im):
"""
对图像进行顶格
:param binary_im: 二值化的图像
:return:
"""
for x in xrange(0, binary_im.shape[0]):
line_val = binary_im[x]
# 不全部是白色(1）
if not (line_val == 1).all():
line_start = x
break
for x1 in xrange(binary_im.shape[0]-1, -1, -1):
line_val = binary_im[x1]
# 不全部是白色(1）
if not (line_val == 1).all():
line_end = x1
break
for y in xrange(0, binary_im.shape[1]):
col_val = binary_im[:, y]
# 不全部是白色(1）
if not (col_val == 1).all():
col_start = y
break
for y1 in xrange(binary_im.shape[1]-1, -1, -1):
col_val = binary_im[:, y1]
# 不全部是白色(1）
if not (col_val == 1).all():
col_end = y1
break
ding_ge_im = binary_im[line_start:line_end, col_start:col_end]
# ding_ge_im = binary_im[:, col_start:col_end]
return ding_ge_im
def make_train_char_db(ori_img_dir):
"""
均分5份，制作训练集
:param ori_img_dir: 原始的图像的目录
:return:
"""
even_split_train_path = os.path.join(os.getcwd(), 'train_data')
if not os.path.exists(even_split_train_path):
os.makedirs(even_split_train_path)
train_imgs = os.listdir(ori_img_dir)
letters = list('0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ')
file = open('label.txt')
answer_data = file.readlines()
# 保存数据
img = np.zeros((len(train_imgs)*4, 1, 32, 32), dtype=np.uint8)
label = np.zeros((len(train_imgs)*4,), dtype='int')
index = 0
for train_img in train_imgs:
if '._' in train_img:
train_img = train_img[2:]
print train_img
ori_train_img = os.path.join(ori_img_dir, train_img)
s_img = cv2.imread(ori_train_img,cv2.IMREAD_GRAYSCALE)
if(s_img is None):
break
#close op
I0 = cv2.morphologyEx(255-s_img, cv2.MORPH_CLOSE, np.ones((5, 5), dtype=np.uint8))
#blackhat op
I1 = cv2.morphologyEx(s_img, cv2.MORPH_BLACKHAT, np.ones((5, 5), dtype=np.uint8))
img_closed = cv2.add(I0, I1)
binary_train_img = binary_img(img_closed) # 二值化之后的图像
binary_train_img = ding_ge(binary_train_img) # 顶格之后的图像
line = answer_data[int(train_img[:-4]) - 1]
print line
# 均分成4份
step_train = binary_train_img.shape[1] / float(4)
start_train = [j for j in np.arange(0, binary_train_img.shape[1], step_train).tolist()]
for p, k in enumerate(start_train):
print train_img + '_' + str((p+1))
split_train_img = binary_train_img[:, k:k + step_train]
small_img = ding_ge(split_train_img)
split_train_resize_img = cv2.resize(small_img, (32, 32))
img[index, 0] = split_train_resize_img
label[index] = letters.index(line[p])
print label[index]
index += 1
#cv2.imwrite(os.path.join(even_split_train_path,
#train_img.split('.')[0] + '_' + str(p+1) + '.png'), #split_train_resize_img*255)
file.close()
caffe_train_path = os.path.join(os.getcwd(), 'captcha_train_lmdb')
env = lmdb.open(caffe_train_path,map_size=500000000)
txn=env.begin(write=True)
count=0
for i in range(img.shape[0]):
datum=caffe.io.array_to_datum(img[i],label[i])
str_id='{:08}'.format(count)
txn.put(str_id,datum.SerializeToString())
count+=1
if count%1000==0:
print('already handled with {} pictures'.format(count))
txn.commit()
txn=env.begin(write=True)
txn.commit()
env.close()
def make_test_char_db(ori_img_dir):
"""
均分5份，制作测试集
:param ori_img_dir: 原始的图像的目录
:return:
"""
even_split_test_path = os.path.join(os.getcwd(), 'test_data')
if not os.path.exists(even_split_test_path):
os.makedirs(even_split_test_path)
test_imgs = os.listdir(ori_img_dir)
# 保存数据
img = np.zeros((len(test_imgs)*4, 1, 32, 32), dtype=np.uint8)
label = np.zeros((len(test_imgs)*4,), dtype='int')
letters = list('0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ')
file = open('5986.txt')
answer_data = file.readlines()
index = 0
for test_img in test_imgs:
if '._' in test_img:
test_img = test_img[2:]
print test_img
ori_test_img = os.path.join(ori_img_dir, test_img)
s_img = cv2.imread(ori_test_img,cv2.IMREAD_GRAYSCALE)
if(s_img is None):
break
#close op
I0 = cv2.morphologyEx(255 - s_img, cv2.MORPH_CLOSE, np.ones((5, 5), dtype='uint8'))
#blackhat op
I1 = cv2.morphologyEx(s_img, cv2.MORPH_BLACKHAT, np.ones((5, 5), dtype='uint8'))
img_closed = cv2.add(I0, I1)
binary_test_img = binary_img(img_closed) # 二值化之后的图像
binary_test_img = ding_ge(binary_test_img) # 顶格之后的图像
line = answer_data[int(test_img[:-4]) - 1]
# 均分成5份
step_test = binary_test_img.shape[1] / float(4)
start_test = [j for j in np.arange(0, binary_test_img.shape[1], step_test).tolist()]
for p, k in enumerate(start_test):
print test_img + '_' + str((p+1))
split_test_img = binary_test_img[:, k:k + step_test]
small_img = ding_ge(split_test_img)
split_test_resize_img = cv2.resize(small_img, (32, 32))
img[index, 0] = split_test_resize_img
label[index] = letters.index(line[p])
index += 1
cv2.imwrite(os.path.join(even_split_test_path,
test_img.split('.')[0] + '_' + str(p+1) + '.png'), split_test_resize_img*255)
file.close()
caffe_test_path = os.path.join(os.getcwd(), 'captcha_test_lmdb')
env = lmdb.open(caffe_test_path,map_size=250000000)
txn=env.begin(write=True)
count=0
for i in range(img.shape[0]):
datum=caffe.io.array_to_datum(img[i],label[i])
str_id='{:08}'.format(count)
txn.put(str_id,datum.SerializeToString())
count+=1
if count%1000==0:
print('already handled with {} pictures'.format(count))
txn.commit()
txn=env.begin(write=True)
txn.commit()
env.close()
if __name__ == '__main__':
#make_train_char_db(os.path.join(os.getcwd(), 'Joker1'))
make_test_char_db(os.path.join(os.getcwd(), 'Joker'))
#make_test_char_db(os.path.join(os.getcwd(), 'test'))