How to reshape an array from vstack into regular stack?-CodePudding

I accidentally used np.vstack(x) instead of np.stack(x, axis=0). Is there a way to reshape that resulting array from vstack into the regular stack(x, axis=0)?

I have the resulting .npy saved on my computer, so if this is possible you just saved me 6 hours of rerunning my code.

Background:

I have 1501 images of size (250,250) that was incorrectly vstacked. The current array shape of these images features is (375250, 2048). I would like this array to be (1501, any number). This is why 375250/250 = 1501.

Each array, before stacking, has shape (2048, )

CodePudding user response：

My computer crashed because of not enough RAM to create an array that big, but theoretically, the following should work:

elements = arr.shape[0] * arr.shape[1]
new_col_num = elements//1501

arr2 = arr.reshape(1501, new_col_num)

arr is the array of size (375250, 2048) and arr2 is the array with shape (1501, some number).

CodePudding user response：

Reshape should be enough, since your intended axis is the first

To illustrate - with a 3d array pretending to be a list of 2 2d arrays:

In [77]: arr = np.arange(24).reshape(2,3,4)
In [78]: np.stack(arr, 0)
Out[78]: 
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]]])
In [79]: np.vstack(arr)
Out[79]: 
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]])
In [80]: np.vstack(arr).reshape(2,3,4)
Out[80]: 
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]]])