在Keras中使用Tensorflow数据集

Question

Tensorflow包tensorflow-datasets有一组广泛的数据集。

我可以执行tfds.list_builders()来查看整个数据集列表，然后使用其中一个数据集如下：

import tensorflow_datasets as tfds
emnist_train = tfds.load(name="emnist", split=tfds.Split.TRAIN, as_supervised=True)
emnist_test  = tfds.load(name="emnist", split=tfds.Split.TEST,  as_supervised=True)

如何使用与emnist_train和emnist_test一起导入的tfds.load和keras load_data()？

Answer 1

您可以像这样加载数据

emnist_train = tfds.load(name="emnist", split=tfds.Split.TRAIN, batch_size=-1 ) 
emnist_test = tfds.load(name="emnist", split=tfds.Split.TEST, batch_size=-1)

emnist_train = tfds.as_numpy(emnist_train) 
emnist_test = tfds.as_numpy(emnist_test)

x_train, y_train = emnist_train["image"], emnist_train["label"] 
x_test, y_test = emnist_test["image"], emnist_test["label"]