keras图像预处理－ImageDataGenerator

相关参数描述：http://keras-cn.readthedocs.io/en/latest/preprocessing/image/
其中validation_split参数（官方上使用方法未描述）：设置训练集与验证集的比例。
　　要与flow_from_directory或flow函数配合。在函数中subset参数中设置为'training' 或者 'validation'，生成对应的数据集。

from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, MaxPooling2D
from keras.models import load_model
import keras.callbacks as callbacks
from keras.preprocessing.image import ImageDataGenerator
 
# 在目录下建立子文件夹，每个子文件夹对应1个类
# 如以0，1，2或a, b, c命名的文件夹
train_dir = 'e:\Python_Project\images\'
 
num_epochs = 1
batch_size = 100
 
data_gen = ImageDataGenerator(rescale=1. / 255, validation_split=0.1)
 
# classes: 可选参数,为子文件夹的列表,如['dogs','cats']默认为None.
# 若未提供,则该类别列表将从directory下的子文件夹名称/结构自动推断。
# 每一个子文件夹都会被认为是一个新的类。(类别的顺序将按照字母表顺序映射到标签值)。
# 通过属性class_indices可获得文件夹名与类的序号的对应字典。
# 本例中使用默认的参数，表示按数字或字母升序，对应类的序号
train_generator = data_gen.flow_from_directory(train_dir,
                                               target_size=(64, 64),
                                               batch_size=batch_size,
                                               class_mode='categorical', subset='training')
validation_generator = data_gen.flow_from_directory(train_dir,
                                               target_size=(64, 64),
                                               batch_size=batch_size,
                                               class_mode='categorical', subset='validation')
# 建立模型
model = Sequential()
# ..
# ...............
 
# 训练模型
tensor_board = callbacks.TensorBoard()
model.fit_generator(generator=train_generator,
                    epochs=num_epochs,
                    validation_data=validation_generator,
                    callbacks=[tensor_board])