Numpy and 2-d array differences

Question

So far, i do what i want to do is as follows

ar2 = [[0 for t in range(maxy-miny)] for t in range(maxx-minx)]

for first,x in enumerate(range(minx, maxx)):
    for second,y in enumerate(range(miny, maxy)):
        ar2[first][second] = a[x][y]

io.imshow(ar2)

The problem is that it's really slow. So i'm considering numpy.

ar = np.zeros((maxx-minx, maxy-miny), dtype=np.ndarray) 

for first,x in enumerate(range(minx, maxx)):
    for second,y in enumerate(range(miny, maxy)):
        ar[first][second] = a[x][y] 

io.imshow(ar)

Though in the second case, the image won't show

TypeError: Image data can not convert to float

I tried to check what might be going on, so i tested ar[0] & ar2[0]

Output of 2-d Array :

[array([230, 197, 204], dtype=uint8) array([241, 209, 214], dtype=uint8)
array([233, 201, 206], dtype=uint8) array([214, 183, 188], dtype=uint8)...

Output of Numpy Array :

[array([230, 197, 204], dtype=uint8), array([241, 209, 214], dtype=uint8),

array([233, 201, 206], dtype=uint8), ...

So apparently, numpy is using commas but i can't understand how and why it happens.

@AndyHayden It's an 2d array of the original image. I have found which parts interest me, and i want to transfer them alone to a new one. — Rentonie
– Rentonie, Commented Dec 20, 2014 at 20:49
As @sebix says it's not a proper 2D numpy array (the dtype is not correct). The spaces vs commas thing is a red herring, that is purely how the objects are printed. — Andy Hayden
– Andy Hayden, Commented Dec 20, 2014 at 20:54
@AndyHayden How can i provide more info? Which part could be done differently? my a is done as follows "a = io.imread('C:\Users\Dimitrios\Desktop\polimesa\\punk.jpg')" — Rentonie
– Rentonie, Commented Dec 20, 2014 at 20:56
I can't promise anything myself, but usually helpful to provide reproducible example (i.e. something we can try at home), in this case this means showing us a or at least the first few elements/rows/columns of a. — Andy Hayden
– Andy Hayden, Commented Dec 20, 2014 at 21:05

sebix · Accepted Answer · 2014-12-20 22:40:42Z

1

Your used dtype is not a valid number-like type, you used np.ndarray as type for individual values in your array.

import matplotlib.pyplot as plt
import numpy as np

ar = np.zeros((maxx-minx, maxy-miny))
a = plt.imread('something.jpg')

for first,x in enumerate(range(minx, maxx)):
    for second,y in enumerate(range(miny, maxy)):
        ar[first][second] = a[x][y] 

plt.imshow(ar)

This should work - however you haven't provided enough information to test that reliably.

I also recommend to use plt as abbreviation for matplotlib.pyplot instead of io, which can easily be mixed up with the actual io-module.

However, you can achieve the same goal but not looping at all:

ar3[0:maxx-minx, 0:maxy-miny] = a[minx:maxx,miny:maxy]
np.allclose(ar2, ar3)  # True

edited Dec 20, 2014 at 22:40

answered Dec 20, 2014 at 20:16

sebix

3,2972 gold badges31 silver badges49 bronze badges

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

2 Comments

Rentonie Over a year ago

i get "ValueError: setting an array element with a sequence" though . I also used "from skimage import io" before.

sebix Over a year ago

I cannot reproduce that. Please provide a fully reproducible code piece. I also added a solution which works without loops.

Collectives™ on Stack Overflow

Numpy and 2-d array differences

1 Answer 1

2 Comments

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Related