numpy: split an array with different start points

Question

Can someone explain to me how to accomplish this?

[
    [1, 2, 9, 10, 17, 18],
    [3, 4, 11, 12, 19, 20],
    [5, 6, 13, 14, 21, 22],
    [7, 8, 15, 16, 23, 24]
]

=>

[
    [
        [1, 2],
        [3, 4]
    ],
    [
        [2, 9],
        [4, 11]
    ],
    [
        [9, 10],
        [11, 12]
    ],
    [
        [10, 17],
        [12, 19]
    ],
    [
        [17, 18],
        [19, 20]
    ]
],
[
    [
        [5, 6],
        [7, 8]
    ],
    [
        [6, 13],
        [8, 15]
    ]
    ...
]

I tried to use numpy.split but this allows only for non-overlapping subarrays.

Basically these are all subarrays of size (2,2) where an imaginary window is moved from top left to bottom right by one column each time.

Chris · Accepted Answer · 2021-11-01 16:57:40Z

As long as you have numpy version 1.20.0 or higher, you can use sliding window view to pass in a window size and get the the form you are looking for.

import numpy as np
from numpy.lib.stride_tricks import sliding_window_view
arr = np.array([
    [1, 2, 9, 10, 17, 18],
    [3, 4, 11, 12, 19, 20],
    [5, 6, 13, 14, 21, 22],
    [7, 8, 15, 16, 23, 24]
])

sliding_window_view(arr,(2,2))

Output

array([[[[ 1,  2],
         [ 3,  4]],

        [[ 2,  9],
         [ 4, 11]],

        [[ 9, 10],
         [11, 12]],

        [[10, 17],
         [12, 19]],

        [[17, 18],
         [19, 20]]],


       [[[ 3,  4],
         [ 5,  6]],

        [[ 4, 11],
         [ 6, 13]],

        [[11, 12],
         [13, 14]],

        [[12, 19],
         [14, 21]],

        [[19, 20],
         [21, 22]]],


       [[[ 5,  6],
         [ 7,  8]],

        [[ 6, 13],
         [ 8, 15]],

        [[13, 14],
         [15, 16]],

        [[14, 21],
         [16, 23]],

        [[21, 22],
         [23, 24]]]])

Collectives™ on Stack Overflow

numpy: split an array with different start points

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related