从 2D 表示 3D 矩阵

Question

我有一个要求，假设您有 5 个文本文件，例如 a、b、c、d、e，其中值以矩阵形式表示，即 a、b、c、d、e 的形状为 (5,5)、(4 ,4)、(7,7)、(6,6)、(8,8)

所以我想要的是在从每个文件中获取矩阵之后，我必须通过在此处填充零将形状转换为 (1,25,25) 1 指的是 index ，就像最后我需要一个形状为 (5, 25,25)

shapes conversion:

(5, 5) -> (1,25,25)
(4, 4) -> (2,25,25)
(7, 7) -> (3,25,25)
(6, 6) -> (4,25,25)
(8, 8) -> (5,25,25)

最后全部放在一起最终形状应该是（5,25,25）

简单示例

a = [[1,2],[3,4]] 

b = [[5,6],[7,8]] 

make a and b in a single list and return like this 
c = [[[1,2],[3,4]],[[5,6],[7,8]]] 

now the shape of c should be (2,2,2)

这是我期望的实际输出

array([[[ 36.85810471,   2.90763259,   2.90761209, ...,   0.        ,
           0.        ,   0.        ],
        [  2.90763259,   0.5       ,   0.29672   , ...,   0.        ,
           0.        ,   0.        ],
        [  2.90761209,   0.29672   ,   0.5       , ...,   0.        ,
           0.        ,   0.        ],
        ..., 
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ]],

       [[ 36.85810471,  12.59994411,   2.90199971, ...,   0.        ,
           0.        ,   0.        ],
        [ 12.59994411,  36.85810471,   1.47311664, ...,   0.        ,
           0.        ,   0.        ],
        [  2.90199971,   1.47311664,   0.5       , ...,   0.        ,
           0.        ,   0.        ],
        ..., 
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ]],

       [[ 36.85810471,  14.26182747,   1.503703  , ...,   0.        ,
           0.        ,   0.        ],
        [ 14.26182747,  36.85810471,   2.92502046, ...,   0.        ,
           0.        ,   0.        ],
        [  1.503703  ,   2.92502046,   0.5       , ...,   0.        ,
           0.        ,   0.        ],
        ..., 
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ]],

       ..., 
       [[ 36.85810471,   8.56999111,  13.29380131, ...,   0.        ,
           0.        ,   0.        ],
        [  8.56999111,  53.35870743,  19.15359688, ...,   0.        ,
           0.        ,   0.        ],
        [ 13.29380131,  19.15359688,  36.85810471, ...,   0.        ,
           0.        ,   0.        ],
        ..., 
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ]],

       [[ 36.85810471,  12.54030132,   8.02613068, ...,   0.        ,
           0.        ,   0.        ],
        [ 12.54030132,  36.85810471,  12.64339542, ...,   0.        ,
           0.        ,   0.        ],
        [  8.02613068,  12.64339542,  36.85810471, ...,   0.        ,
           0.        ,   0.        ],
        ..., 
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ]],

       [[ 36.85810471,  12.62930584,  12.60999584, ...,   0.        ,
           0.        ,   0.        ],
        [ 12.62930584,  36.85810471,   7.73449707, ...,   0.        ,
           0.        ,   0.        ],
        [ 12.60999584,   7.73449707,  36.85810471, ...,   0.        ,
           0.        ,   0.        ],
        ..., 
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ],
        [  0.        ,   0.        ,   0.        , ...,   0.        ,
           0.        ,   0.        ]]], dtype=float32)

形状是 (7165,23,23)

谁能告诉我如何做到这一点？

Answer 1

这是您可以做我认为您想做的一种方法。 假设您的初始数组是a 、 b等，例如：

a = np.arange(25).reshape((5,5))
b = np.arange(36).reshape((6,6))
c = np.arange(16).reshape((4,4))
...

然后填充它们并堆叠它们：

W = np.dstack([np.pad(m,((0,25-m.shape[0]),(0,25-m.shape[0])),
                      mode='constant') for m in (a,b,c)])
X = np.rollaxis(W, 2)

X.shape然后是(3, 25, 25)原始矩阵条目在每个“层”的左上角。 您需要滚动轴，因为深度堆叠它们会得到一个形状为(25, 25, 3)的数组。