train.py

from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D
from keras.optimizers import SGD
from keras.callbacks import EarlyStopping
import numpy as np

X_train = np.load("X_train.npy")
Y_train = np.load("Y_train.npy")

batch_size = 32
nb_classes = 62
nb_epoch = 400
data_augmentation = True

# input image dimensions
img_rows, img_cols = 22, 15
img_channels = 3

print('X_train shape:', X_train.shape)
print(X_train.shape[0], 'train samples')

model = Sequential()

model.add(Convolution2D(32, 3, 3, border_mode='same',
                        input_shape=X_train.shape[1:]))
model.add(Activation('relu'))
model.add(Convolution2D(32, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))

model.add(Convolution2D(64, 3, 3, border_mode='same'))
model.add(Activation('relu'))
model.add(Convolution2D(64, 3, 3))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))

model.add(Flatten())
model.add(Dense(512))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes))
model.add(Activation('softmax'))

sgd = SGD(lr=1e-5, decay=0, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy',
              optimizer=sgd)

model.fit(X_train, Y_train, batch_size=batch_size, nb_epoch=nb_epoch,
          verbose=1)
# score = model.evaluate(X_test, Y_test, verbose=0)
# print('Test score:', score[0])
# print('Test accuracy:', score[1])

model.save_weights('shu_captcha_CNN_weights.h5')
json_string = model.to_json()
f=open('shu_captcha_CNN_structure.json','w')
f.write(json_string)
f.close()