numpy.ndarray.transpose

一、ndarray.transpose的官方文档：

ndarray.transpose(*axes)
Returns a view of the array with axes transposed.
//返回对array进行维度转置后的结果

For a 1-D array this has no effect, as a transposed vector is simply the same vector.
//一维array转置后仍然是其本身

To convert a 1-D array into a 2D column vector, an additional dimension must be added. np.atleast2d(a).T achieves this, as does a[:, np.newaxis].

For a 2-D array, this is a standard matrix transpose.
//对二维array，进行标准矩阵转置

For an n-D array, if axes are given, their order indicates how the axes are permuted (see Examples). If axes are not provided and a.shape = (i[0], i[1], … i[n-2], i[n-1]), then a.transpose().shape = (i[n-1], i[n-2], … i[1], i[0]).

Parameters
axes：None, tuple of ints, or n ints
None or no argument: reverses the order of the axes.

tuple of ints: i in the j-th place in the tuple means a’s i-th axis becomes a.transpose()’s j-th axis.
//将原array中第i个维度置于新array的第j个维度

n ints: same as an n-tuple of the same ints (this form is intended simply as a “convenience” alternative to the tuple form)

Returns：out：ndarray
View of a, with axes suitably permuted.

官方文档中给的Examples：

a = np.array([[1, 2], [3, 4]])
a

array([[1, 2],
[3, 4]])
//使用np.array()创建array

a.transpose()

array([[1, 3],
[2, 4]])
//本Example作用：说明文档中“If axes are not provided and a.shape = (i[0], i[1], … i[n-2], i[n-1]), then a.transpose().shape = (i[n-1], i[n-2], … i[1], i[0]).”

a.transpose((1, 0))

array([[1, 3],
[2, 4]])
//本Example作用：说明文档中“tuple of ints: i in the j-th place in the tuple means a’s i-th axis becomes a.transpose()’s j-th axis.”的参数形式

a.transpose(1, 0)

array([[1, 3],
[2, 4]])
//本Example作用：说明文档中“n ints: same as an n-tuple of the same ints (this form is intended simply as a “convenience” alternative to the tuple form)”参数形式

注意的是：由于上述函数没有将转置后的结果进行返回，因此array a始终是最初的形式

二、使用ndarray.transpose函数：
（1）原array：

In:a = np.array([[[3,6],[2,7],[8,6]],[[4,3],[5,8],[0,3]]])
In:a
Out[6]: 
array([[[3, 6],
        [2, 7],
        [8, 6]],
       [[4, 3],
        [5, 8],
        [0, 3]]])
In:a.shape
Out[7]: (2, 3, 2)

（2）进行“If axes are not provided and a.shape = (i[0], i[1], … i[n-2], i[n-1]), then a.transpose().shape = (i[n-1], i[n-2], … i[1], i[0]).”形式的转置：

In:a.transpose()
Out[8]: 
array([[[3, 4],
        [2, 5],
        [8, 0]],
       [[6, 3],
        [7, 8],
        [6, 3]]])
In:a.transpose().shape
Out[10]: (2, 3, 2)

此处为三维矩阵进行转置，如下图所示：

[1]https://segmentfault.com/a/1190000016731276

沿着绿色方向线进行交换
因为与绿色方向线垂直的方向是一维轴线和三维轴线的角平分线

（3）进行“uple of ints: i in the j-th place in the tuple means a’s i-th axis becomes a.transpose()’s j-th axis.”形式的转置：

In：a.transpose((1,2,0))
Out[11]: 
array([[[3, 4],
        [6, 3]],
       [[2, 5],
        [7, 8]],
       [[8, 0],
        [6, 3]]])

转置直观图如下所示：

在（2）中，该转置操作改变了立方体的构造；
在（3）中，该转置操作仅仅改变了立方体的摆放位置，但是没有改变构造。

三、注意事项：
1、无order参数，则默认使用“全部颠倒过来”（If axes are not provided and a.shape = (i[0], i[1], … i[n-2], i[n-1]), then a.transpose().shape = (i[n-1], i[n-2], … i[1], i[0]).）的置换方式
2、order参数可以是int tuple（如a.transpose((2,0,1))），也可以是ints（如a.transpose(2,0,1)，注意，不是tuple）

四、使用场景：

_img = cv2.imread(filename)
_img = cv2.resize(_img, (mycfg.IMG_WIDTH, mycfg.IMG_HEIGHT), interpolation=cv2.INTER_CUBIC)
_img = _img.transpose((2, 0, 1))
_img = _img / 255
_img = torch.from_numpy(_img)
_img = _img.to(torch.float32)
return _img

使用cv2.imread()函数读入img，形式为numpy.ndarray，其shape为(height,weight,channel)
进行转置，变换为(channel,height,weight)

使用a.transpose(2,0,1)并没有改变原array（如上图所示立方体）的任何构造，只是进行了位置改变
因此图片内容（各个像素之间的位置关系）没有发生改变

猜测：此处对图像进行transpose(2,0,1)转置操作，是为了_img = torch.from_numpy(_img)，将其转为张量。待更新。