从图像numpy生成一批克隆 [英] generating batch of clones from image numpy
本文介绍了从图像numpy生成一批克隆的处理方法,对大家解决问题具有一定的参考价值,需要的朋友们下面随着小编来一起学习吧!
问题描述
我有一个名为 a 的 numpy 数组(图像),其大小为:
I have a numpy array (an image) called a with this size:
[3,128,192]
现在我想创建一个包含 a 个副本的numpy数组,该副本具有以下维度:
now i want create a numpy array that contains n copies of a which will have this dimension:
[n,3,128,192]
存在一个numpy函数,可以在不使用循环指令的情况下帮助我解决此问题?
exist a numpy function that can help me with this problem without using loop instructions?
推荐答案
只需使用np.stack
# say you need 10 copies of a 3D array `a`
In [267]: n = 10
In [266]: np.stack([a]*n)
或者,如果您确实担心性能,则应该使用np.concatenate
.
Alternatively, you should use np.concatenate
if you're really concerned about the performance.
In [285]: np.concatenate([a[np.newaxis, :, :]]*n)
示例:
In [268]: a
Out[268]:
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]]])
In [271]: a.shape
Out[271]: (3, 4, 4)
In [269]: n = 10
In [270]: np.stack([a]*n).shape
Out[270]: (10, 3, 4, 4)
In [285]: np.concatenate([a[np.newaxis, :, :]]*n).shape
Out[285]: (10, 3, 4, 4)
性能:
Performance:
# ~ 4x faster than using `np.stack`
In [292]: %timeit np.concatenate([a[np.newaxis, :, :]]*n)
100000 loops, best of 3: 10.7 µs per loop
In [293]: %timeit np.stack([a]*n)
10000 loops, best of 3: 41.1 µs per loop
这篇关于从图像numpy生成一批克隆的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!
查看全文