-
Notifications
You must be signed in to change notification settings - Fork 0
/
image_ocr_train.py
511 lines (445 loc) · 20 KB
/
image_ocr_train.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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
# -*- coding: utf-8 -*-
'''CNN+GRU+CTC network for optical character recognition of generated characters images.
This source code bases on "image_ocr.py" in keras examples.
Some main changes are listed as follows:
* Generate characters in random rather than loading words from file.
* Change network architecture. BN is added after
Conv2D and one BiGRU is removed, since the origin network
is found hard to converge.
* Reduce epochs since BN accelerates training.
* The function "on_train_begin" seems working in parallel with
"fit_generator", so built word list on constructor.
* Replace cairocffi with PIL. cairocffi relies on GTK which is not easy
to install on windows.
* If font size is too large to paint, try smaller size rather than
throw a exception immediately.
* Rotation range is tuned to avoid string exceeding canvas.
* Support saving "predict_model".
The origin [keras example](https://github.com/keras-team/keras/blob/master/examples/image_ocr.py)
was created by [Mike Henry](https://github.com/mbhenry/)
'''
import os
# os.environ["CUDA_VISIBLE_DEVICES"]="-1"
import itertools
import codecs
import re
import datetime
import numpy as np
from scipy import ndimage
import pylab
from PIL import Image, ImageDraw, ImageFont
from tensorflow.python.keras import backend as K
from tensorflow.python.keras.layers import Conv2D, MaxPooling2D, BatchNormalization
from tensorflow.python.keras.layers import Input, Dense, Activation
from tensorflow.python.keras.layers import Reshape, Lambda
from tensorflow.python.keras.layers import add, concatenate
from tensorflow.python.keras.models import Model
from tensorflow.python.keras.layers import GRU
from tensorflow.python.keras.optimizers import SGD
from tensorflow.python.keras.preprocessing import image
from tensorflow.python.keras import callbacks
OUTPUT_DIR = './keras/ocr/image_ocr'
# digit classes
# alphabet = u'0123456789 '
# English characters classes
alphabet = u'abcdefghijklmnopqrstuvwxyz '
# this creates larger "blotches" of noise which look
# more realistic than just adding gaussian noise
# assumes greyscale with pixels ranging from 0 to 1
def speckle(img):
severity = np.random.uniform(0, 0.6)
blur = ndimage.gaussian_filter(np.random.randn(*img.shape) * severity, 1)
img_speck = (img + blur)
img_speck[img_speck > 1] = 1
img_speck[img_speck <= 0] = 0
return img_speck
# paints the string in a random location the bounding box
# also uses a random font, a slight random rotation,
# and a random amount of speckle noise
def paint_text(text, w, h, rotate=False, ud=False, multi_fonts=False):
surface = Image.new('L', (w, h), 255) # "L" represents gray
border_w_h = (2, 2)
box_width = 0
box_height = 0
for font_size in range(32, 20, -1):
if multi_fonts:
fonts = ["simsun.ttc", "simhei.ttf", "msyh.ttc", "msyhbd.ttc", "msyhl.ttc",
"simfang.ttf", "simkai.ttf"]
font = ImageFont.truetype(np.random.choice(fonts), font_size)
else:
font = ImageFont.truetype("msyh.ttc", font_size)
box_width, box_height = font.getsize(text)
if box_width <= (w - 2 * border_w_h[0]) and box_height <= (h - border_w_h[1]):
break
elif font_size <= 21:
raise IOError(('Could not fit string into image.'
'Max char count is too large for given image width.'))
# teach the RNN translational invariance by
# fitting text box randomly on canvas, with some room to rotate
max_shift_x = w - box_width - border_w_h[0]
max_shift_y = h - box_height - border_w_h[1]
top_left_x = np.random.randint(border_w_h[0], int(max_shift_x) + 1)
if ud:
top_left_y = np.random.randint(0, int(max_shift_y) + 1)
else:
top_left_y = max_shift_y // 2
draw = ImageDraw.Draw(surface)
draw.text((top_left_x, top_left_y), text, fill=0, font=font)
a = np.array(surface, np.uint8)
a = a.astype(np.float32) / 255
a = np.expand_dims(a, 0)
if rotate: # avoid string exceeding canvas
a = image.random_rotation(a, 70 * (np.min([top_left_y, h - box_height - top_left_y])) / w)
a = speckle(a)
return a
def shuffle_mats_or_lists(matrix_list, stop_ind=None):
ret = []
assert all([len(i) == len(matrix_list[0]) for i in matrix_list])
len_val = len(matrix_list[0])
if stop_ind is None:
stop_ind = len_val
assert stop_ind <= len_val
a = list(range(stop_ind))
np.random.shuffle(a)
a += list(range(stop_ind, len_val))
for mat in matrix_list:
if isinstance(mat, np.ndarray):
ret.append(mat[a])
elif isinstance(mat, list):
ret.append([mat[i] for i in a])
else:
raise TypeError('`shuffle_mats_or_lists` only supports '
'numpy.array and list objects.')
return ret
# Translation of characters to unique integer values
def text_to_labels(text):
ret = []
for char in text:
ret.append(alphabet.find(char))
return ret
# Reverse translation of numerical classes back to characters
def labels_to_text(labels):
ret = []
for c in labels:
if c == len(alphabet): # CTC Blank
ret.append("")
else:
ret.append(alphabet[c])
return "".join(ret)
# Uses generator functions to supply train/test with
# data. Image renderings are text are created on the fly
# each time with random perturbations
class TextImageGenerator(callbacks.Callback):
def __init__(self, minibatch_size, img_w, img_h,
downsample_factor, val_split, build_word_count,
max_string_len=25, mono_fraction=1):
self.minibatch_size = minibatch_size
self.img_w = img_w
self.img_h = img_h
self.downsample_factor = downsample_factor
self.val_split = val_split
self.blank_label = self.get_output_size() - 1
self.max_string_len = max_string_len
self.mono_fraction = mono_fraction
self.cur_val_index = 0
self.cur_train_index = 0
self.build_word_list(build_word_count)
self.paint_func = lambda text: paint_text(
text, self.img_w, self.img_h,
rotate=False, ud=False, multi_fonts=False)
def get_output_size(self):
return len(alphabet) + 1
# num_words can be independent of the epoch size due to the use of generators
# as max_string_len grows, num_words can grow
def build_word_list(self, num_words):
assert num_words % self.minibatch_size == 0
assert (self.val_split * num_words) % self.minibatch_size == 0
self.num_words = num_words
self.string_list = [''] * self.num_words
tmp_string_list = []
self.Y_data = np.ones([self.num_words, self.max_string_len]) * -1
self.X_text = []
self.Y_len = [0] * self.num_words
def _is_length_of_word_valid(word):
return (self.max_string_len == -1 or
self.max_string_len is None or
len(word) <= self.max_string_len)
# generate characters in random
for _ in range(int(self.num_words * self.mono_fraction)):
word_len = np.random.randint(self.max_string_len // 2, self.max_string_len + 1)
word = ""
for _ in range(word_len):
word += alphabet[np.random.randint(0, len(alphabet) - 1)]
if _is_length_of_word_valid(word):
tmp_string_list.append(word)
# generate characters with black in random (seems hard to train)
for _ in range(self.num_words - int(self.num_words * self.mono_fraction)):
word_len = np.random.randint(self.max_string_len // 2, self.max_string_len + 1)
word = ""
for _ in range(word_len):
word += alphabet[np.random.randint(0, len(alphabet))]
if _is_length_of_word_valid(word):
tmp_string_list.append(word)
if len(tmp_string_list) != self.num_words:
raise IOError('Could not pull enough words'
'from supplied monogram and bigram files.')
# interlace to mix up the easy and hard words
self.string_list[::2] = tmp_string_list[:self.num_words // 2]
self.string_list[1::2] = tmp_string_list[self.num_words // 2:]
for i, word in enumerate(self.string_list):
self.Y_len[i] = len(word)
self.Y_data[i, 0:len(word)] = text_to_labels(word)
self.X_text.append(word)
self.Y_len = np.expand_dims(np.array(self.Y_len), 1)
self.cur_val_index = self.val_split
self.cur_train_index = 0
# each time an image is requested from train/val/test, a new random
# painting of the text is performed
def get_batch(self, index, size, train):
# width and height are backwards from typical Keras convention
# because width is the time dimension when it gets fed into the RNN
if K.image_data_format() == 'channels_first':
X_data = np.ones([size, 1, self.img_w, self.img_h])
else:
X_data = np.ones([size, self.img_w, self.img_h, 1])
labels = np.ones([size, self.max_string_len])
input_length = np.zeros([size, 1])
label_length = np.zeros([size, 1])
source_str = []
for i in range(size):
# Mix in some blank inputs. This seems to be important for
# achieving translational invariance
if train and i > size - 2:
if K.image_data_format() == 'channels_first':
X_data[i, 0, 0:self.img_w, :] = self.paint_func('')[0, :, :].T
else:
X_data[i, 0:self.img_w, :, 0] = self.paint_func('',)[0, :, :].T
labels[i, 0] = self.blank_label
input_length[i] = self.img_w // self.downsample_factor - 2
label_length[i] = 1
source_str.append('')
else:
if K.image_data_format() == 'channels_first':
X_data[i, 0, 0:self.img_w, :] = (
self.paint_func(self.X_text[index + i])[0, :, :].T)
else:
X_data[i, 0:self.img_w, :, 0] = (
self.paint_func(self.X_text[index + i])[0, :, :].T)
labels[i, :] = self.Y_data[index + i]
input_length[i] = self.img_w // self.downsample_factor - 2
label_length[i] = self.Y_len[index + i]
source_str.append(self.X_text[index + i])
inputs = {'the_input': X_data,
'the_labels': labels,
'input_length': input_length,
'label_length': label_length,
'source_str': source_str # used for visualization only
}
outputs = {'ctc': np.zeros([size])} # dummy data for dummy loss function
return (inputs, outputs)
def next_train(self):
while 1:
ret = self.get_batch(self.cur_train_index,
self.minibatch_size, train=True)
self.cur_train_index += self.minibatch_size
if self.cur_train_index >= self.val_split:
self.cur_train_index = self.cur_train_index % 32
(self.X_text, self.Y_data, self.Y_len) = shuffle_mats_or_lists(
[self.X_text, self.Y_data, self.Y_len], self.val_split)
yield ret
def next_val(self):
while 1:
ret = self.get_batch(self.cur_val_index,
self.minibatch_size, train=False)
self.cur_val_index += self.minibatch_size
if self.cur_val_index >= self.num_words:
self.cur_val_index = self.val_split + self.cur_val_index % 32
yield ret
def on_epoch_begin(self, epoch, logs={}):
# rebind the paint function to implement curriculum learning
if 2 <= epoch < 4:
self.paint_func = lambda text: paint_text(
text, self.img_w, self.img_h,
rotate=False, ud=True, multi_fonts=False)
elif 4 <= epoch < 6:
self.paint_func = lambda text: paint_text(
text, self.img_w, self.img_h,
rotate=False, ud=True, multi_fonts=True)
elif epoch >= 6:
self.paint_func = lambda text: paint_text(
text, self.img_w, self.img_h,
rotate=True, ud=True, multi_fonts=True)
# the actual loss calc occurs here despite it not being
# an internal Keras loss function
def ctc_lambda_func(args):
y_pred, labels, input_length, label_length = args
# the 2 is critical here since the first couple outputs of the RNN
# tend to be garbage:
y_pred = y_pred[:, 2:, :]
return K.ctc_batch_cost(labels, y_pred, input_length, label_length)
# For a real OCR application, this should be beam search with a dictionary
# and language model. For this example, best path is sufficient.
def decode_batch(test_func, word_batch):
out = test_func([word_batch])[0]
ret = []
for j in range(out.shape[0]):
out_best = list(np.argmax(out[j, 2:], 1))
out_best = [k for k, g in itertools.groupby(out_best)]
outstr = labels_to_text(out_best)
ret.append(outstr)
return ret
class VizCallback(callbacks.Callback):
def __init__(self, run_name, test_func, text_img_gen, num_display_words=6):
self.test_func = test_func
self.output_dir = os.path.join(
OUTPUT_DIR, run_name)
self.text_img_gen = text_img_gen
self.num_display_words = num_display_words
if not os.path.exists(self.output_dir):
os.makedirs(self.output_dir)
def on_epoch_end(self, epoch, logs={}):
weight_file = os.path.join(self.output_dir, 'weights%02d.h5' % (epoch))
self.model.save_weights(weight_file)
print("save_weight: ", weight_file)
word_batch = next(self.text_img_gen)[0]
res = decode_batch(self.test_func,
word_batch['the_input'][0:self.num_display_words])
if word_batch['the_input'][0].shape[0] < 256:
cols = 2
else:
cols = 1
for i in range(self.num_display_words):
pylab.subplot(self.num_display_words // cols, cols, i + 1)
if K.image_data_format() == 'channels_first':
the_input = word_batch['the_input'][i, 0, :, :]
else:
the_input = word_batch['the_input'][i, :, :, 0]
pylab.imshow(the_input.T, cmap='Greys_r')
pylab.xlabel(
'Truth = \'{}\'\nPredict = \'{}\' ({})'
.format(word_batch['source_str'][i], res[i], word_batch['source_str'][i] == res[i]))
fig = pylab.gcf()
fig.set_size_inches(10, 13)
pylab.savefig(os.path.join(self.output_dir, 'e%02d.png' % (epoch)))
pylab.close()
def train(run_name, start_epoch, stop_epoch, img_w, build_word_count,
max_string_len, mono_fraction, save_model_path=None):
# Input Parameters
img_h = 64
words_per_epoch = 16000
val_split = 0.2
val_words = int(words_per_epoch * (val_split))
# Network parameters
conv_filters = 16
kernel_size = (3, 3)
pool_size = 2
time_dense_size = 32
# GRU output NAN when rnn_size=512 with my GPU, but CPU or rnn_size=256 is ok.
# Tensorflow 1.10 appears, but vanishes in 1.12!
rnn_size = 512
minibatch_size = 32
# if start_epoch >= 12:
# minibatch_size = 8 # 32 is to large for my poor GPU
if K.image_data_format() == 'channels_first':
input_shape = (1, img_w, img_h)
else:
input_shape = (img_w, img_h, 1)
img_gen = TextImageGenerator(
minibatch_size=minibatch_size,
img_w=img_w,
img_h=img_h,
downsample_factor=(pool_size ** 2),
val_split=words_per_epoch - val_words,
build_word_count=build_word_count,
max_string_len=max_string_len,
mono_fraction=mono_fraction)
act = 'relu'
kernel_init = 'he_normal'
input_data = Input(name='the_input', shape=input_shape, dtype='float32')
inner = Conv2D(conv_filters, kernel_size, padding='same',
activation=None, kernel_initializer=kernel_init,
name='conv1')(input_data)
inner = BatchNormalization(axis=3, scale=False, name='bn1')(inner)
inner = Activation(activation=act)(inner)
inner = MaxPooling2D(pool_size=(pool_size, pool_size), name='max1')(inner)
inner = Conv2D(conv_filters, kernel_size, padding='same',
activation=None, kernel_initializer=kernel_init,
name='conv2')(inner)
inner = BatchNormalization(axis=3, scale=False, name='bn2')(inner)
inner = Activation(activation=act)(inner)
inner = MaxPooling2D(pool_size=(pool_size, pool_size), name='max2')(inner)
conv_to_rnn_dims = (img_w // (pool_size ** 2),
(img_h // (pool_size ** 2)) * conv_filters)
inner = Reshape(target_shape=conv_to_rnn_dims, name='reshape')(inner)
# cuts down input size going into RNN:
inner = Dense(time_dense_size, activation=act, name='dense1')(inner)
# bidirectional GRU, GRU seems to work as well, if not better than LSTM:
gru_1 = GRU(rnn_size, return_sequences=True,
kernel_initializer=kernel_init, name='gru1')(inner)
gru_1b = GRU(rnn_size, return_sequences=True,
go_backwards=True, kernel_initializer=kernel_init,
name='gru1_b')(inner)
gru1_merged = concatenate([gru_1, gru_1b])
# transforms RNN output to character activations:
inner = Dense(img_gen.get_output_size(), kernel_initializer=kernel_init,
name='dense2')(gru1_merged)
y_pred = Activation('softmax', name='softmax')(inner)
labels = Input(name='the_labels',
shape=[img_gen.max_string_len], dtype='float32')
input_length = Input(name='input_length', shape=[1], dtype='int64')
label_length = Input(name='label_length', shape=[1], dtype='int64')
# Keras doesn't currently support loss funcs with extra parameters
# so CTC loss is implemented in a lambda layer
loss_out = Lambda(
ctc_lambda_func, output_shape=(1,),
name='ctc')([y_pred, labels, input_length, label_length])
# clipnorm seems to speeds up convergence
sgd = SGD(lr=0.02, decay=1e-6, momentum=0.9, nesterov=True, clipnorm=5)
model = Model(inputs=[input_data, labels, input_length, label_length],
outputs=loss_out)
# the loss calc occurs elsewhere, so use a dummy lambda func for the loss
model.compile(loss={'ctc': lambda y_true, y_pred: y_pred},
optimizer=sgd,
metrics=['accuracy'])
if start_epoch > 0:
weight_file = os.path.join(
OUTPUT_DIR,
os.path.join(run_name, 'weights%02d.h5' % (start_epoch - 1)))
model.load_weights(weight_file)
print("load_weight: ", weight_file)
# captures output of softmax so we can decode the output during visualization
test_func = K.function([input_data], [y_pred])
viz_cb = VizCallback(run_name, test_func, img_gen.next_val())
model.fit_generator(
generator=img_gen.next_train(),
steps_per_epoch=(words_per_epoch - val_words) // minibatch_size,
epochs=stop_epoch,
validation_data=img_gen.next_val(),
validation_steps=val_words // minibatch_size,
callbacks=[viz_cb, img_gen],
initial_epoch=start_epoch,
verbose=1)
if save_model_path:
predict_model = Model(inputs=input_data, outputs=y_pred)
predict_model.save(save_model_path)
if __name__ == '__main__':
RUN_NAME = datetime.datetime.now().strftime('%Y-%m-%d_%H-%M-%S')
# RUN_NAME = "EnglishWord_GRU"
SAVE_MODEL_PATH = "./model/image_ocr_word.h5"
train(run_name=RUN_NAME,
start_epoch=0,
stop_epoch=12,
img_w=128,
build_word_count=16000,
max_string_len=5,
mono_fraction=1)
# increase to wider images and start at epoch 12.
# The learned weights are reloaded
train(run_name=RUN_NAME,
start_epoch=12,
stop_epoch=20,
img_w=512,
build_word_count=32000,
max_string_len=25,
mono_fraction=1,
save_model_path=SAVE_MODEL_PATH)