What does this third argument of reshape mean?

Question

I was reading this question about numpy, array and reshape. I understand what the OP is doing and his line of code, except the last argument of reshape: 3

According to the documentation of numpy.reshape, the third argument of the function is the order and should be a string {'C', 'F', 'A'}, optional.

So what does the OP argument 3 means?

Side note: In the question, OP places 3 as second argument because it is the function numpy.array.reshape, while in the documentation of numpy.reshape the order is the third argument. But it is because in numpy.reshape the array itself is the first argument.

The numpy.array.reshape documentation page redirects to numpy.reshape page.

sshashank124 · Accepted Answer · 2020-01-12T19:09:16.917

2

You are looking at the wrong version of reshape. The relevant one is ndarray.reshape which

allows the elements of the shape parameter to be passed in as separate arguments

Therefore,

a = np.array(...)
a.reshape(3, 4, 5)

is like doing

np.reshape(a, (3, 4, 5))

In the original question, the 3 is simply a part of the reshape operation since the OP is trying to work with an RGB image as a 3D array of shape (height, width, 3)

edited Jan 12 '20 at 19:09

answered Jan 12 '20 at 17:35

sshashank124

31,495
9
67
76

In the side note of the question, I do explain that I noticed that, except that I have called "ndarray.reshape" as "nupy.array.reshape" – totalMongot Jan 12 '20 at 18:39
The `3` is just part of specifying the shape for the reshape operation – sshashank124 Jan 12 '20 at 18:41
But the shape has already been explained in the example by len(X) and len(Y). What does the parameter 3 mean ? – totalMongot Jan 12 '20 at 18:47
They are trying to reshape the data for a color image so the 3 corresponds to the depth of the array/image i.e. the RGB channels. – sshashank124 Jan 12 '20 at 18:48
Ok so that mean that three arrays are handled for one picture, ie one array for each of the RGB channels? – totalMongot Jan 12 '20 at 18:56
Better to think of it as 1 array which is of shape (height, width, 3), so essentially 3 single color frames stacked on top of each other – sshashank124 Jan 12 '20 at 18:57
Ok I get it Maybe you can give an answer with this discussion of comments and I will accept it? – totalMongot Jan 12 '20 at 19:06

score 0 · Answer 2 · answered Oct 28 '20 at 18:05

Consider this code, which runs as expected:

arr = np.arange(12) # arr has 12 element
arr = np.reshape(arr, (3, 4)) 

print(arr) # prints '[[ 0  1  2  3] [ 4  5  6  7] [ 8  9 10 11]]'

Now, consider this code which has an error in it:

arr = np.arange(24) # arr has 24 element

# next line will fail, because (3, 4) can only have 12 element... but 'arr' has 24 element
arr = np.reshape(arr, (3, 4)) 

print(arr)

What can we do to solve it?

A solution would be, to have 3 parents list inside the main list, where each parent list has a structure of (4, 2). So together these 3 parents list can hold 8 * 3 or 24 items. That's why the next block of code runs without any error.

arr = np.arange(24) # arr has 24 element

arr = np.reshape(arr, (3, 4, 2))  # 3 means it will have 3 parent list that has a structure of (4, 2)

print(arr) # prints '[  [[ 0  1] [ 2  3] [ 4  5] [ 6  7]]    [[ 8  9] [10 11] [12 13] [14 15]]    [[16 17] [18 19] [20 21] [22 23]]  ]'

What does this third argument of reshape mean?

2 Answers2