# 先进行预处理图像
train_datagen = ImageDataGenerator(rescale=1./255, 
                  rotation_range=50,
                  height_shift_range=[-0.005, 0, 0.005],
                  width_shift_range=[-0.005, 0, 0.005],
                  horizontal_flip=True, 
                  fill_mode='reflect')
#再对预处理图像指定从目录中读取数据，可以看到我的目录最核心的地方是images_keras_dict（可以对照上一张图片）
train_generator = train_datagen.flow_from_directory('AgriculturalDisease_trainingset/images_keras_dict',
                          target_size=(height, width), batch_size=16)

val_datagen = ImageDataGenerator(rescale=1./255)
val_generator = val_datagen.flow_from_directory('AgriculturalDisease_validationset/images_keras_dict', target_size=(height, width),
                        batch_size=64)

save_weights = ModelCheckpoint(filepath='models/best_weights.hdf5',monitor='val_loss', verbose=1, save_best_only=True)

# 最后在fit_generator 中放入生成器的函数train_generator
model.fit_generator(train_generator,
          steps_per_epoch=times_train,
          verbose=1,
          epochs=300,
          initial_epoch=0,
          validation_data=val_generator,
          validation_steps=times_val,
          callbacks=[save_weights, TrainValTensorBoard(write_graph=False)])

# 建立预处理
predict_datagen = ImageDataGenerator(rescale=1./255)
predict_generator = predict_datagen.flow_from_directory('AgriculturalDisease_validationset/test',
                            target_size=(height, width), batch_size=128)
# predict_generator.reset()
# 利用predict_generator进行预测
pred = model.predict_generator(predict_generator, max_queue_size=10, workers=1, verbose=1)

# 利用几个属性来读取文件夹和对应的分类
train_datagen = ImageDataGenerator(rescale=1./255, rotation_range=40, fill_mode='wrap')
train_generator = train_datagen.flow_from_directory('new_images', target_size=(height, width), batch_size=96)
labels = (train_generator.class_indices)
labels = dict((v,k) for k,v in labels.items())
predictions = [labels[k] for k in predicted_class_indices]

# 还可以知道图片的名字
filenames = predict_generator.filenames

import tensorflow as tf
model = tf.keras.models.Sequential([
 ... // 你的模型
])
model.fit(train_x, // 训练输入
  train_y, // 训练标签
  epochs=5 // 训练5轮
)

def generator(x, y, b_size):
 ... // 处理函数

model.fit_generator(generator(train_x, train_y, batch_size), 
   step_per_epochs=np.ceil(len(train_x)/batch_size), 
   epochs=5
)

def generator(x, y, b_size):
 """Generates batch and batch and batch then feed into models.
 Args:
 x: input data;
 y: input labels;
 b_size: batch_size.
 Yield:
 (batch_x, batch_label): batched x and y.
 """
 while 1: // 死循环
 idx = ...
 batch_x = ...
 batch_y = ...
 ... // 任何你想要对这个`batch`中的数据执行的操作
 yield (batch_x, batch_y)

class Generator(keras.utils.Sequence):
 def __init__(self, x, y, b_size):
 self.x, self.y = x, y
 self.batch_size = b_size
 
 def __len__(self):
 return math.ceil(len(self.y)/self.batch_size

 def __getitem__(self, idx):
 b_x = self.x[idx*self.batch_size:(idx+1)*self.batch_size]
 b_y = self.y[idx*self.batch_size:(idx+1)*self.batch_size]
 ... // 对`batch`的其余操作
 return np.array(b_x), np.array(b_y)
 
 def on_epoch_end(self):
 """执行完一个`epoch`之后，还可以做一些其他的事情！"""
 ...

train_generator = Generator(train_x, train_y, batch_size)
val_generator = Generator(val_x, val_y, batch_size)

model.fit_generator(generator=train_generator, 
   epochs=3197747, 
   validation_data=val_generator
   )