tf.keras.preprocessing.image.random_shift

Performs a random spatial shift of a Numpy image tensor.

tf.keras.preprocessing.image.random_shift(
    x,
    wrg,
    hrg,
    row_axis=1,
    col_axis=2,
    channel_axis=0,
    fill_mode='nearest',
    cval=0.0,
    interpolation_order=1
)

x Input tensor. Must be 3D. wrg Width shift range, as a float fraction of the width. hrg Height shift range, as a float fraction of the height. row_axis Index of axis for rows in the input tensor. col_axis Index of axis for columns in the input tensor. channel_axis Index of axis for channels in the input tensor. fill_mode Points outside the boundaries of the input are filled according to the given mode (one of {'constant', 'nearest', 'reflect', 'wrap'}). cval Value used for points outside the boundaries of the input if mode='constant'. interpolation_order int, order of spline interpolation. see ndimage.interpolation.affine_transform

Shifted Numpy image tensor.