How to reshape an array from vstack into regular stack?

Question

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, )

Answer 1

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]]])

Answer 2

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).