I would like to convert numpy array into numpy array of arrays.
I have an array: a = [[0,0,0],[0,255,0],[0,255,255],[255,255,255]]
and I would like to have: b = [[[0,0,0],[0,0,0],[0,0,0]],[[0,0,0],[255,255,255],[0,0,0]],[[0,0,0],[255,255,255],[255,255,255]],[[255,255,255],[255,255,255],[255,255,255]]]
Is there any easy way to do it?
I have tried with np.where(a == 0, [0,0,0],[255,255,255]) but I got the following error:
ValueError: operands could not be broadcast together with shapes
You can use broadcast_to as
b = np.broadcast_to(a, (3,4,3))
where a was shape (3,4). Then you need to swap the axes around
import numpy as np
a = np.array([[0,0,0],[0,255,0],[0,255,255],[255,255,255]])
b = np.broadcast_to(a, (3,4,3))
c = np.moveaxis(b, [0,1,2], [2,0,1])
c
giving
array([[[ 0, 0, 0],
[ 0, 0, 0],
[ 0, 0, 0]],
[[ 0, 0, 0],
[255, 255, 255],
[ 0, 0, 0]],
[[ 0, 0, 0],
[255, 255, 255],
[255, 255, 255]],
[[255, 255, 255],
[255, 255, 255],
[255, 255, 255]]])
A more direct method broadcasting method suggested by @Divakar is
b = np.broadcast(a[:,:,None], (4,3,3))
which produces the same output without axis swapping.
np.broadcast_to(a[:,:,None], (4,3,3)).What you tried will work with the following small modification:
a = np.array(a)
np.where(a[...,None]==0,[0,0,0],[255,255,255])
To make multidimensional indexing available we have to cast a to array first. a[...,None] adds a new dimension at the end of a to accomodate the triplets 0,0,0 and 255,255,255.
In [204]: a = np.array([[0,0,0],[0,255,0],[0,255,255],[255,255,255]])
In [205]: a.shape
Out[205]: (4, 3)
Looks like you want to replicate each element 3 times, making a new trailing dimension. We can do that using repeat (after adding the new trailing dimension):
In [207]: a.reshape(4,3,1).repeat(3,2)
Out[207]:
array([[[ 0, 0, 0],
[ 0, 0, 0],
[ 0, 0, 0]],
[[ 0, 0, 0],
[255, 255, 255],
[ 0, 0, 0]],
[[ 0, 0, 0],
[255, 255, 255],
[255, 255, 255]],
[[255, 255, 255],
[255, 255, 255],
[255, 255, 255]]])
In [208]: _.shape
Out[208]: (4, 3, 3)
