Can anybody explain me the numpy.indices()?

Suppose you have a matrix M whose (i,j)-th element equals

M_ij = 2*i + 3*j

One way to define this matrix would be

i, j = np.indices((2,3))M = 2*i + 3*j

which yields

array([[0, 3, 6],       [2, 5, 8]])

In other words, np.indices returns arrays which can be used as indices. The elements in i indicate the row index:

In [12]: iOut[12]: array([[0, 0, 0],       [1, 1, 1]])

The elements in j indicate the column index:

In [13]: jOut[13]: array([[0, 1, 2],       [0, 1, 2]])

The already posted answers are still complex so here a the simplest way to understand this.

Step 1: Let's create a 2x2 grid

ids = np.indices((2,2))

Step 2: Now let's unpack the i,j indices

i, j = ids 

These are the indices i,j:

print(i)[[0 0] [1 1]]print(j)[[0 1] [0 1]]

Step 3: Understand what i,j represent

The easy way to think of it is to make pairs as (i0,j0), (i1,j1), (i2,j2), (i3,j3) i.e. match each element of i with the corresponding element of j.

So we get: (0,0), (0,1), (1,0), (1,1).

These are exactly the indices of a 2x2 grid:

I've understood with this code.

The following function has the same behavior as np.indices().

# fixed dimensions=(2,3,4)def my_indices():    dimensions = (2,3,4)    A = np.empty(dimensions)    # dimensions[0] = 2    A[0, :, :] = 0    A[1, :, :] = 1    B = np.empty(dimensions)    # dimensions[1] = 3    B[:, 0, :] = 0    B[:, 1, :] = 1    B[:, 2, :] = 2    C = np.empty(dimensions)    # dimensions[2] = 4    C[:, :, 0] = 0    C[:, :, 1] = 1    C[:, :, 2] = 2    C[:, :, 3] = 3    return [A, B, C] 

Call

A, B, C = my_indices()print(A.shape)print(B.shape)print(C.shape)print('A\n', A)print('B\n', B)print('C\n', C)

RESULT

(2, 3, 4)(2, 3, 4)(2, 3, 4)A [[[0. 0. 0. 0.]  [0. 0. 0. 0.]  [0. 0. 0. 0.]] [[1. 1. 1. 1.]  [1. 1. 1. 1.]  [1. 1. 1. 1.]]]B [[[0. 0. 0. 0.]  [1. 1. 1. 1.]  [2. 2. 2. 2.]] [[0. 0. 0. 0.]  [1. 1. 1. 1.]  [2. 2. 2. 2.]]]C [[[0. 1. 2. 3.]  [0. 1. 2. 3.]  [0. 1. 2. 3.]] [[0. 1. 2. 3.]  [0. 1. 2. 3.]  [0. 1. 2. 3.]]]

np.indices() use case

def create_hsv_map():    img_hsv = np.empty((180, 256, 3), np.uint8)    hue, saturation = np.indices((180,256))    img_hsv[:, :, 0] = hue    img_hsv[:, :, 1] = saturation    img_hsv[:, :, 2] = 255    # ...

example with np.repeat() instead of np.indices()

def create_hsv_map2():    img_hsv = np.empty((180, 256, 3), np.uint8)    hue = np.repeat(np.arange(180).reshape(180, 1), repeats=256, axis=1)    saturation = np.repeat(np.arange(256).reshape(1, 256), repeats=180, axis=0)    img_hsv[:, :, 0] = hue    img_hsv[:, :, 1] = saturation    img_hsv[:, :, 2] = 255    # ...