BatchDataset 转换为 Numpy 数组

Question

#parameter
batch_size = 32
img_height = 48
img_width = 48
#load image
train_ds = tf.keras.preprocessing.image_dataset_from_directory(
    'Lampung_PNG',
    label_mode = "int", #categorical or binary
    #class_name=['0', '1', '2', '3', ...]
    color_mode='grayscale',
    batch_size=batch_size,
    image_size=(img_height, img_width),
    shuffle=True,
    seed=9898,
    validation_split=0.1,
    subset="training",
)
valid_ds = tf.keras.preprocessing.image_dataset_from_directory(
    'Lampung_PNG',
    label_mode = "int", #categorical or binary
    #class_name=['0', '1', '2', '3', ...]
    color_mode='grayscale',
    batch_size=batch_size,
    image_size=(img_height, img_width),
    shuffle=True,
    seed=9898,
    validation_split=0.1,
    subset="validation",
)

valid_ds =

我的问题 = 如何将 train_ds 和 valid_ds 转换为 numpy 数组？我想应用这个，但我不能使用我当前的数据集。

sm = SVMSMOTE(random_state=42)
x_res, y_res = sm.fit_resample(train_ds, valid_ds)
print('Resampled dataset shape %s' % Counter(y_res))

Answer 1

您找到解决方案了吗？如果你能分享，我会很高兴。