How to check values in columns in a numpy array

Question

I have a numpy array,

> [[  0   0   0   0   0]
 [  0   0   0   0   0]
 [  0   0   0   0   0]
 ...
 [  0 255   0 255   0]
 [  0 255   0   0   0]
 [  0 255   0   0   0]]

I would like to check values in each column of this 2d numpy array - whether or not there are more than one column with alternating 0 and 255 values

Example column

>   [[  0 0   0 255   0]
     [  0 255 0   0   0]
     [  0 255 0   0   0]
     [  0 255 0   0   0]
     [  0 0   0   0   0]
     [  0 0   0   0   0]
     [  0 0   0   0   0]
     [  0 0   0   0   0]
     [  0 255 0   0   0]
     [  0 255 0   0   0]
     [  0 255 0   0   0]
     [  0 255 0   0   0]
     [  0 255 0   0   0]]

In the above example, column two has alternating 0 and 255 values, hence, the output should be that this numpy array has one column with these alternating values

The fourth column also has alternating 0 and 255 values, no? — Lev Zakharov
– Lev Zakharov, Commented Aug 7, 2018 at 15:00
It has, but I will add a threshold of the occurences of 255 and 0s — Label Lane det
– Label Lane det, Commented Aug 9, 2018 at 7:32

Pulkit Maloo · Accepted Answer · 2018-08-07 14:57:05Z

1

np.sum(np.count_nonzero(np.diff(a, axis=0), axis=0) > 1)

edited Aug 7, 2018 at 14:57

answered Aug 7, 2018 at 14:51

Pulkit Maloo

112 bronze badges

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

3 Comments

andrew_reece Over a year ago

Wow, beautiful.

Label Lane det Over a year ago

Does it check all the columns of the array?

Label Lane det Over a year ago

I need the number of columns that have a majority of alternating 0s and 255s.

Krishna Chouhan · Accepted Answer · 2018-08-07 14:52:19Z

0

I dont understand how the second column has alternate values of 0 and 255 but here is an idea...

>>> unique, counts = numpy.unique(numpy.transpose(nd_a)[0], return_counts=True)
>>> counts
array([13], dtype=int64)
>>> unique, counts = numpy.unique(numpy.transpose(nd_a)[1], return_counts=True)
>>> counts
array([5, 8], dtype=int64)

if len(nd_a) is same as counts then its not alternating but if values are somewhat similar(e.g. 5 and 8 take 5/13 and 8/13) then its alternating.

answered Aug 7, 2018 at 14:52

Krishna Chouhan

991 silver badge8 bronze badges

Comments

andrew_reece · Accepted Answer · 2018-08-07 15:02:13Z

Assuming "alternating" means that the same value appears more than once without being adjacent to its other occurrences:

If you take the diff of each row's value with its successor (per column), there should be three possible values: 0, 255, -255. Only columns with alternating values will have both 255 and -255. Testing for the presence of these two diff values (using a few Pandas methods) will give you the answer.

import pandas as pd

def check_alt(data):
    return (pd.DataFrame(data)
            .diff()
            .apply(lambda x: x.value_counts())
            .loc[[-255,255]]
            .notnull()
            .all()
            .sum())

check_alt(a) # 2
check_alt(b) # 4

Data:

a = [[0,0,0,255,0],
[0,255,0,0,255],
[0,255,0,0,0],
[0,255,0,0,0],
[0,0,0,0,0],
[0,0,0,0,0],
[0,0,0,0,0],
[0,0,0,0,0],
[0,255,0,0,0],
[0,255,0,0,0],
[0,255,0,0,0],
[0,255,0,0,0],
[0,255,0,0,0]]

b = [[0,0,0,255,0],
[255,255,0,0,255],
[0,0,0,255,0]]

The nested list a is your example data. I also included an extra array, b, to demonstrate the edge case where there are only constantly alternating values (i.e. every next value is not the one before it). The expected count of alternating columns for b is 4.

Andrew Palmer · Accepted Answer · 2018-08-07 15:15:52Z

Here is another solution;

import numpy as np

#construct numpy array
c = np.zeros((15,5)) # Pre-allocate matrix


#populate some columns with number patterns

c[1:4,1] = 255
c[8:13,1] = 255

c[1,3] = 255

c[1:4,4] = 255
c[7:12,4] = 255

OUT = np.empty(shape=[15,0])

for ni in range(0,c.shape[1]):
    #work on column ni from c
    a = c[:,ni] #work on array a
    b = np.diff(a) #differene between neighbouring elements
    d = np.where(b != 0) #work out where the transitions from 0->255 or 255->0 are

    #work out what constitutes an alternating pattern (you may want to change the logic)
    if np.size(d) > 2:
        #this column from C has been deemed to have an alternating pattern
        OUT = np.append(OUT, c[:,ni:ni+1], axis=1) # add the column to the OUT array

Collectives™ on Stack Overflow

How to check values in columns in a numpy array

4 Answers 4

3 Comments

Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

3 Comments

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related