Question

我有兴趣用随机图像变换来扩充我的数据集。我正在使用Keras ImageDataGenerator，并且在尝试将random_transform应用于单个图片时出现以下错误：

--> x = apply_transform(x, transform matrix, img_channel_axis, fill_mode, cval)
>>> RuntimeError: affine matrix has wrong number of rows.

我找到了ImageDataGenerator here的源代码。但是，我不确定如何调试运行时错误。以下是我的代码：

from keras.preprocessing.image import img_to_array, load_img
from keras.preprocessing.image import ImageDataGenerator
from keras.applications.inception_v3 import preprocess_input    

image_path = './figures/zebra.jpg'

#data augmentation
train_datagen = ImageDataGenerator(
    rotation_range=40,
    width_shift_range=0.2,
    height_shift_range=0.2,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=True,
    fill_mode='nearest')

print "\nloading image..."
image = load_img(image_path, target_size=(299, 299))
image = img_to_array(image)
image = np.expand_dims(image, axis=0) # 1 x input_shape
image = preprocess_input(image)

train_datagen.fit(image)
image = train_datagen.random_transform(image)

调用random_transform时，错误发生在最后一行。

Answer 1

问题在于random_transform需要3D阵列。

请参阅docstring：

def random_transform(self, x, seed=None):
    """Randomly augment a single image tensor.
    # Arguments
        x: 3D tensor, single image.
        seed: random seed.
    # Returns
        A randomly transformed version of the input (same shape).
    """

因此，您需要在np.expand_dims之前调用它。

Keras ImageDataGenerator：随机变换

1 个答案: