Python - Converting 3D numpy array to 2D

Question

I have a 3D matrix like this:

array([[[ 0,  1],
      [ 2,  3]],

      [[ 4,  5],
      [ 6,  7]],

      [[ 8,  9],
      [10, 11]],

      [[12, 13],
      [14, 15]]])

and would like to stack them in a grid format, ending up with:

 array([[ 0,  1],
        [ 2,  3],
        [ 4,  5],
        [ 6,  7],
        [ 8,  9],
        [10, 11],
        [12, 13],
        [14, 15]])

Currently, I'm using numpy as a library.

GrozaiL · Accepted Answer · 2018-10-03 15:22:45Z

9

I suggest you to visit this link also for this case would work np.reshape((-1,2))

answered Oct 3, 2018 at 15:16

GrozaiL

3743 silver badges6 bronze badges

1

$\begingroup$ This works nicely! the exactly syntax would be arr.reshape((-1, 2)), and the link seems to be wrong, reshape is docs.scipy.org/doc/numpy-1.15.0/reference/generated/…. @GrozaiL, but this (-1, 2) feels like magic for me, could you explain how it works? Thanks! $\endgroup$
– Indominus
Commented Jan 14, 2019 at 3:56

Add a comment |

Rama · Accepted Answer · 2020-01-29 15:51:28Z

The numpy.reshape() allows you to do reshaping in multiple ways.

It usually unravels the array row by row and then reshapes to the way you want it.

If you want it to unravel the array in column order you need to use the argument order='F'

Let's say the array is a. For the case above, you have a (4, 2, 2) ndarray

numpy.reshape(a, (8, 2)) will work.

In the general case of a (l, m, n) ndarray:

numpy.reshape(a, (l*m, n)) should be used.

Or, numpy.reshape(a, (a.shape[0]*a.shape[1], a.shape[2])) should be used.

2 Answers 2