Python Numpy中的reshape与transpose的用法以及区别_python中reshape和transpose的区别

今天在处理图像时，发现输入图像的尺寸为（长、宽、通道数），需要改为（通道数、长、宽），起初打算使用np.reshape想当然的修改，但却发现已经修改了图像的内容。后来使用np.transpose修改成功。在此记录下两个函数的区别。

np.reshape

在Numpy中，reshape的操作首先将多维数组转为一维向量，再按照转变后新数组的维度，截取拼成新的形状。一定不要用来处理图片信息！！！

下面希望将(2,4,3)形状的图像，转变成通道数3在前面的(3,2,4)的图像。

>>> import numpy as np
>>> x = np.random.rand(2,4,3)*10
>>> x_reshape = x.reshape((3,2,4))
>>> x
array([[[8.03765126, 4.39127424, 2.74543376],
        [9.37186554, 1.23974842, 0.7722333 ],
        [5.84788233, 5.74285694, 2.08277486],
        [4.12951592, 2.93530013, 6.5999786 ]],

       [[6.32425653, 8.86223448, 7.46397057],
        [2.16678125, 2.20945112, 1.25329685],
        [2.5462106 , 8.10238692, 3.1542975 ],
        [7.56363603, 9.06540176, 4.72479797]]])
>>> x_reshape
array([[[8.03765126, 4.39127424, 2.74543376, 9.37186554],
        [1.23974842, 0.7722333 , 5.84788233, 5.74285694]],

       [[2.08277486, 4.12951592, 2.93530013, 6.5999786 ],
        [6.32425653, 8.86223448, 7.46397057, 2.16678125]],

       [[2.20945112, 1.25329685, 2.5462106 , 8.10238692],
        [3.1542975 , 7.56363603, 9.06540176, 4.72479797]]])
#	相当于先转到1维再截取切分
>>> x.reshape(1, 24)
array([[8.03765126, 4.39127424, 2.74543376, 9.37186554, 1.23974842,
        0.7722333 , 5.84788233, 5.74285694, 2.08277486, 4.12951592,
        2.93530013, 6.5999786 , 6.32425653, 8.86223448, 7.46397057,
        2.16678125, 2.20945112, 1.25329685, 2.5462106 , 8.10238692,
        3.1542975 , 7.56363603, 9.06540176, 4.72479797]])
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

上面就是把一个(2,4,3)形状的数组，先变成一维数组，然后在reshape成(3,2,4)的形状

np.transpose

>>> x_transpose = x.transpose(2,0,1)
>>> x
array([[[8.03765126, 4.39127424, 2.74543376],
        [9.37186554, 1.23974842, 0.7722333 ],
        [5.84788233, 5.74285694, 2.08277486],
        [4.12951592, 2.93530013, 6.5999786 ]],

       [[6.32425653, 8.86223448, 7.46397057],
        [2.16678125, 2.20945112, 1.25329685],
        [2.5462106 , 8.10238692, 3.1542975 ],
        [7.56363603, 9.06540176, 4.72479797]]])
>>> x_transpose
array([[[8.03765126, 9.37186554, 5.84788233, 4.12951592],
        [6.32425653, 2.16678125, 2.5462106 , 7.56363603]],

       [[4.39127424, 1.23974842, 5.74285694, 2.93530013],
        [8.86223448, 2.20945112, 8.10238692, 9.06540176]],

       [[2.74543376, 0.7722333 , 2.08277486, 6.5999786 ],
        [7.46397057, 1.25329685, 3.1542975 , 4.72479797]]])
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20