forked from tuandoan998/Handwritten-Text-Recognition
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Train.py
58 lines (48 loc) · 2.08 KB
/
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
from Parameter import *
from ImageGenerator import TextImageGenerator
from CRNN_Model import word_model, line_model
from keras.callbacks import ModelCheckpoint, EarlyStopping
from keras.optimizers import Adam
from Utils import *
def train(train_data, val_data, is_word_model):
if is_word_model:
model, _ = word_model()
cfg = word_cfg
else:
model, _ = line_model()
cfg = line_cfg
input_length = cfg['input_length']
model_name = cfg['model_name']
max_text_len = cfg['max_text_len']
img_w = cfg['img_w']
img_h = cfg['img_h']
batch_size = cfg['batch_size']
train_set = TextImageGenerator(train_data, img_w, img_h, batch_size, input_length, max_text_len)
print('Loading data for train ...')
train_set.build_data()
val_set = TextImageGenerator(val_data, img_w, img_h, batch_size, input_length, max_text_len)
val_set.build_data()
print('Done')
print("Number train samples: ", train_set.n)
print("Number val samples: ", val_set.n)
model.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer='adam')
ckp = ModelCheckpoint(
filepath='Resource/'+model_name+'--{epoch:02d}--{val_loss:.3f}.h5', monitor='val_loss',
verbose=1, save_best_only=True, save_weights_only=True
)
earlystop = EarlyStopping(
monitor='val_loss', min_delta=0, patience=10, verbose=0, mode='min'
)
model.fit_generator(generator=train_set.next_batch(),
steps_per_epoch=train_set.n // batch_size,
epochs=32,
validation_data=val_set.next_batch(),
validation_steps=val_set.n // batch_size,
callbacks=[ckp, earlystop])
return model
if __name__=='__main__':
train_data = get_paths_and_texts('data/IAM/splits/train.uttlist', is_words=True)
val_data = get_paths_and_texts('data/IAM/splits/validation.uttlist', is_words=True)
print('number of train image: ', len(train_data))
print('number of valid image: ', len(val_data))
model = train(train_data, val_data, True)