/
character_recognizer.py
executable file
·238 lines (197 loc) · 9.34 KB
/
character_recognizer.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
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
'''
Created on 08/07/2015
@author: Alexandre Yukio Yamashita
Flavio Nicastro
'''
from ConfigParser import SafeConfigParser
from argparse import ArgumentParser
from curses.ascii import isalpha
import cv2
from numpy.core.defchararray import isdigit
from models.image import Image
from models.logger import Logger
import numpy as np
class CharacterRecognizer:
'''
Recognize character from binary image.
'''
def __init__(self, character_width=15, character_height=15, classifier_type="svm_linear"):
self.character_width = character_width
self.character_height = character_height
self.classifier_type = classifier_type
self.resize = 128
if classifier_type == "knn":
self.classifier = cv2.KNearest()
else:
self.classifier = cv2.SVM()
if self.classifier_type == "svm_linear_hog":
winSize = (128,128)
blockSize = (32,32)
blockStride = (16,16)
cellSize = (8,8)
nbins = 9
derivAperture = 1
winSigma = 4.
histogramNormType = 0
L2HysThreshold = 2.0000000000000001e-01
gammaCorrection = 0
nlevels = 32
self.hog = cv2.HOGDescriptor(winSize,blockSize,blockStride,cellSize,nbins,derivAperture,winSigma,
histogramNormType,L2HysThreshold,gammaCorrection,nlevels)
def _pre_process(self, image):
if self.classifier_type == "svm_linear_hog":
image = image.resize(self.resize, self.resize)
image.invert_binary()
image.filter_gaussian_blur(size=49)
#image.plot()
else:
image = image.resize(self.character_width, self.character_height)
image.invert_binary()
return image
def train(self, samples, responses):
'''
Train classifier.
'''
# Get binaries to train
binaries = []
for sample in samples:
image = Image(image=sample.data)
image = self._pre_process(image)
if self.classifier_type == "svm_linear_hog":
hog_feature = self.hog.compute(image.data)
binaries.append(hog_feature)
else:
binaries.append(image.resize(self.character_width, self.character_height))
if self.classifier_type != "svm_linear_hog":
binaries = np.array([np.array(image.data.flatten(), dtype=np.float32) for image in binaries])
else:
binaries = np.array([np.array(image.flatten(), dtype=np.float32) for image in binaries])
# Train classifier.
responses = np.array(responses)
if self.classifier_type == "knn":
self.samples = binaries
self.labels = responses
self.classifier.train(binaries, responses)
elif self.classifier_type == "svm_rbf":
params = dict(kernel_type=cv2.SVM_RBF,
svm_type=cv2.SVM_C_SVC,
C=1)
self.classifier.train(binaries, responses, params=params)
else:
params = dict(kernel_type=cv2.SVM_LINEAR,
svm_type=cv2.SVM_C_SVC,
C=1)
self.classifier.train(binaries, responses, params=params)
def load(self, trained_classifier_file, path_images="", path_labels=""):
'''
Load configuration from file.
'''
if self.classifier_type != "knn":
self.classifier.load(trained_classifier_file)
else:
self.samples = np.loadtxt(path_images, np.float32)
self.labels = np.loadtxt(path_labels, np.float32)
self.classifier.train(self.samples, self.labels)
def predict(self, sample, is_letter=False):
'''
Predict sample label.
'''
image = Image(image=sample.data)
image = self._pre_process(image)
if self.classifier_type == "svm_linear_hog":
image = self.hog.compute(image.data)
image = np.array(image.flatten(), dtype=np.float32)
else:
image = np.array(image.data.flatten(), dtype=np.float32)
if self.classifier_type != "knn":
label = self.classifier.predict(image)
else:
_, results, _, _ = self.classifier.find_nearest(np.array([image]), k = 1)
label = results[0][0]
label = chr(int(label))
return label
def save(self, trained_classifier_file, path_images="", path_labels=""):
'''
Save configuration.
'''
if self.classifier_type != "knn":
self.classifier.save(trained_classifier_file)
else:
np.savetxt(path_images, self.samples)
np.savetxt(path_labels, self.labels)
if __name__ == '__main__':
'''
Train letter and number recognizers.
'''
# Parses args.
arg_parser = ArgumentParser(description='Train character recognizer.')
arg_parser.add_argument('-c', '--config', dest='config_file', default='config.ini', help='Configuration file')
args = vars(arg_parser.parse_args())
# Parses configuration file.
config_parser = SafeConfigParser()
config_parser.read(args['config_file'])
character_width = int(config_parser.get('training', 'character_width'))
character_height = int(config_parser.get('training', 'character_height'))
character_original_width = int(config_parser.get('training', 'character_original_width'))
character_original_height = int(config_parser.get('training', 'character_original_height'))
path_letter_images = config_parser.get('data', 'path_letter_images')
path_letter_labels = config_parser.get('data', 'path_letter_labels')
path_letter_classifier = config_parser.get('data', 'path_letter_classifier')
path_number_images = config_parser.get('data', 'path_number_images')
path_number_labels = config_parser.get('data', 'path_number_labels')
path_number_classifier = config_parser.get('data', 'path_number_classifier')
path_number_knn_labels_classifier = config_parser.get('data', 'path_number_knn_labels_classifier')
path_number_knn_images_classifier = config_parser.get('data', 'path_number_knn_images_classifier')
path_letter_knn_labels_classifier = config_parser.get('data', 'path_letter_knn_labels_classifier')
path_letter_knn_images_classifier = config_parser.get('data', 'path_letter_knn_images_classifier')
character_classifier_type = config_parser.get('data', 'character_classifier_type')
# Load images and labels.
logger = Logger()
logger.log(Logger.INFO, "Loading images to train classifiers.")
number_images = np.loadtxt(path_number_images, np.uint8)
number_labels = np.loadtxt(path_number_labels, np.float32)
number_labels = number_labels.reshape((number_labels.size, 1))
converted_images = []
labels = []
for index in range(len(number_images)):
image = number_images[index]
reshaped = Image(image=image.reshape((character_original_height, character_original_width)))
reshaped.binarize(adaptative=True)
mean_value = np.mean(reshaped.data)
if mean_value < 220:
if isdigit(chr(int(number_labels[index]))):
converted_images.append(reshaped)
labels.append(number_labels[index])
number_images = converted_images
number_labels = labels
letter_images = np.loadtxt(path_letter_images, np.uint8)
letter_labels = np.loadtxt(path_letter_labels, np.float32)
letter_labels = letter_labels.reshape((letter_labels.size, 1))
converted_images_l = []
labels_l = []
for index in range(len(letter_images)):
image = letter_images[index]
reshaped = Image(image=image.reshape((character_original_height, character_original_width)))
reshaped.binarize(adaptative=True)
mean_value = np.mean(reshaped.data)
if mean_value < 220:
if isalpha(chr(int(letter_labels[index]))):
converted_images_l.append(reshaped)
labels_l.append(letter_labels[index])
letter_images = converted_images_l
letter_labels = labels_l
# Train classifiers.
logger.log(Logger.INFO, "Training letter classifier.")
letter_classifier = CharacterRecognizer(character_width, character_height, character_classifier_type)
letter_classifier.train(letter_images, letter_labels)
if letter_classifier.classifier_type != "knn":
letter_classifier.save(path_letter_classifier)
else:
letter_classifier.save("", path_letter_knn_images_classifier, path_letter_knn_labels_classifier)
logger.log(Logger.INFO, "Training number classifier.")
number_classifier = CharacterRecognizer(character_width, character_height, character_classifier_type)
number_classifier.train(number_images, number_labels)
if number_classifier.classifier_type != "knn":
number_classifier.save(path_number_classifier)
else:
number_classifier.save("", path_number_knn_images_classifier, path_number_knn_labels_classifier)