tensorflow ImageDataGenerator的使用

#类：train, test都需要有rescale
train_datagen = ImageDataGenerator(
    rotation_range=15,
    rescale=1./255,
    shear_range=0.1,
    zoom_range=0.2,
    horizontal_flip=True,
    width_shift_range=0.1,
    height_shift_range=0.1
)

train_generator = train_datagen.flow_from_dataframe(
    train_df, 
    "../working/train/", 
    x_col='filename',
    y_col='category',
    target_size=IMAGE_SIZE,
    class_mode='categorical',
    batch_size=batch_size
)
#相当于一个迭代器，可以不断生成新的不同的图片
#Pandas dataframe containing the filepaths relative to directory 
#string, column in dataframe that contains the filenames (or absolute paths if directory is None).
#string or list, column/s in dataframe that has the target data.
example_df = train_df.sample(n=1).reset_index(drop=True)
example_generator = train_datagen.flow_from_dataframe(
    example_df, 
    "../working/train/", 
    x_col='filename',
    y_col='category',
    target_size=IMAGE_SIZE,
    class_mode='categorical'
)
plt.figure(figsize=(12, 12))
#生成15张图片(生成的数量根据自己的需要决定)
for i in range(0, 15):
	#subplot: 在i+1的位置上画图
    plt.subplot(5, 3, i+1)
    #generator: 无限的以一个batch为单位生成数据，所以需要取第一个
    #图片，并且在生成第一张图片后break
    for X_batch, Y_batch in example_generator:
        image = X_batch[0]
        plt.imshow(image)
        break
plt.tight_layout()
plt.show()

total_train = train_df.shape[0]
total_validate = validate_df.shape[0]
epochs=3 if FAST_RUN else 50
#由于generator会无限生成数据，所以需要根据steps_per_epoch的值
#决定这个epoch何时结束，何时开始新的epoch
history = model.fit_generator(
    train_generator, 
    epochs=epochs,
    validation_data=validation_generator,
    validation_steps=total_validate//batch_size,
    steps_per_epoch=total_train//batch_size,
    callbacks=callbacks
)

predict = model.predict_generator(test_generator, steps=np.ceil(nb_samples/batch_size))

docs
link
link