Keras ImageDataGenerator.flow_from_directory 返回 TypeError答案

【问题标题】：Keras ImageDataGenerator.flow_from_directory returns TypeErrorKeras ImageDataGenerator.flow_from_directory 返回 TypeError
【发布时间】：2019-11-09 04:33:30
【问题描述】：

我正在尝试将目录作为 ImageDataGenerator.flow_from_directory 的输入，但我无法做到。

train_data_dir = "/train"
validation_data_dir = "/test"

train_generator = ImageDataGenerator.flow_from_directory(directory=train_data_dir,
target_size = (img_height, img_width),
batch_size = batch_size, 
class_mode = "categorical")

validation_generator = ImageDataGenerator.flow_from_directory(directory=validation_data_dir,
target_size = (img_height, img_width),
class_mode = "categorical")

以上代码返回如下错误

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-126-39ed634afa51> in <module>
      2 target_size = (img_height, img_width),
      3 batch_size = batch_size,
----> 4 class_mode = "categorical")
      5 
      6 validation_generator = ImageDataGenerator.flow_from_directory(validation_data_dir,

TypeError: flow_from_directory() missing 1 required positional argument: 'self'

我该如何解决这个问题？

【问题讨论】：

标签： python machine-learning keras deep-learning image-preprocessing

【解决方案1】：

您不能直接从 ImageDataGenerator 调用 flow_from_directory 方法。您必须首先创建此类的实例。试试这个：

train_gen = ImageDataGenerator()
val_gen = ImageDataGenerator()

您可以在此处添加增强参数。参考：https://keras.io/preprocessing/image/
之后，您可以使用 flow_from_directory。

train_generator = train_gen.flow_from_directory(directory=train_data_dir,
target_size = (img_height, img_width),
batch_size = batch_size, 
class_mode = "categorical")

【讨论】：