随机翻转numpy数组中每个图像的最有效方法

时间:2016-04-19 10:54:31

标签: python numpy image-processing deep-learning

例如,我有一个形状为a的numpy数组1000*3*256*256。 换句话说,a是一个包含1000个图像的数组,每个图像的大小为3*256*256。 我想随意翻转每一张图片,所以我的问题是如何有效地做到这一点?谢谢!

1 个答案:

答案 0 :(得分:1)

基础:array[slice(a,b,c)]相当于array[a:b:c],反转(“翻转”)数组使用slice(None, None, -1),这与array[::-1]相同。

因此,让我们为每张图片构建随机翻转:

>>> import random
>> flips = [(slice(None, None, None),
...          slice(None, None, random.choice([-1, None])),
...          slice(None, None, random.choice([-1, None])))
...          for _ in xrange(a.shape[0])]

第一个切片用于通道,第二个切片用于Y轴,第三个切片用于X轴。让我们构建一些测试数据:

>>> import numpy as np
>>> a = np.array(range(3*2*5*5)).reshape(3,2,5,5)

我们可以将每个随机翻转单独应用于每个图像:

>>> flips[0]
(slice(None, None, None), slice(None, None, -1), slice(None, None, None))
>>> a[0]
array([[[ 0,  1,  2,  3,  4],
        [ 5,  6,  7,  8,  9],
        [10, 11, 12, 13, 14],
        [15, 16, 17, 18, 19],
        [20, 21, 22, 23, 24]],

       [[25, 26, 27, 28, 29],
        [30, 31, 32, 33, 34],
        [35, 36, 37, 38, 39],
        [40, 41, 42, 43, 44],
        [45, 46, 47, 48, 49]]])
>>> a[0][flips[0]]
array([[[20, 21, 22, 23, 24],
        [15, 16, 17, 18, 19],
        [10, 11, 12, 13, 14],
        [ 5,  6,  7,  8,  9],
        [ 0,  1,  2,  3,  4]],

       [[45, 46, 47, 48, 49],
        [40, 41, 42, 43, 44],
        [35, 36, 37, 38, 39],
        [30, 31, 32, 33, 34],
        [25, 26, 27, 28, 29]]])

如您所见,flips[0]垂直翻转图像。现在对每张图片都很简单:

>>> random_flipped = np.array([img[flip] for img, flip in zip(a, flips)])