Reshape 3-D numpy.array to 1-D array in specific order

Question

Reshape 3-D numpy array to 1-D array:

I would like to reshape a 3-D array that looks like this:

test_3d = np.array([[[0., 0.],
        [1., 1.]],

       [[2., 2.],
        [3., 3.]]])

To a 1-D array that looks like this:

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

Flattening the array using test_3d.flatten() outputs:

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

A combination of np.flatten/ravel and transpose functionalities work well for 2-D arrays, but for a 3-D array I get the following:

Input:
test_3d.T.flatten()
Output:
array([0., 2., 1., 3., 0., 2., 1., 3.])

Does anybody have any ideas?

transpose takes an order parameter.. Read its docs and experiment. — hpaulj
– hpaulj, Commented May 8, 2021 at 14:59

Valdi_Bo · Accepted Answer · 2021-05-08 17:01:56Z

1

To present a more instructive example, I defined test_3d as:

test_3d = np.array(
    [[[0., 10.],
      [1., 11.]],
     [[2., 12.],
      [3., 13.]]])

(now you can tell apart both "initial" zeroes).

To get your expected result, run:

result = np.transpose(test_3d, (2, 0, 1)).flatten()

The result is:

array([ 0.,  1.,  2.,  3., 10., 11., 12., 13.])

An alternative solution is:

result = np.rollaxis(test_3d, 2).flatten()

answered May 8, 2021 at 17:01

Valdi_Bo

31.1k4 gold badges29 silver badges45 bronze badges

Sign up to request clarification or add additional context in comments.

1 Answer 1