Filter 3D python list using list or numpy

Question

I have a python 3D list. Making it more clear, a list of list where each list is the four corner coordinate of a box. I have to filter all the boxes which are smaller than some given size.

Let us suppose this is the python list.

box = [[[4, 4], [4, 8], [8, 8], [8, 4]],
      [[8, 8], [8, 16], [16, 16], [16, 8]],
      [[20,16],[20,20],[24,20],[24,16]]
      ...
      ]

I have to filter all boxes with length and breadth less than or equal to 5.

filtered_box = [[[4, 4], [4, 8], [8, 8], [8, 4]],
                [[20,16],[20,20],[24,20],[24,16]]
               ...
               ]

This is my current code

filtered_box = []
for c in box:
    min_x, min_y = c[0]
    max_x, max_y = c[2]
    if max_x - min_x <= 5 & min_y - max_y <= 5:
        filtered_box.append(c)

This is working good but I need a more optimized solution. It can use numpy and convert back to python list or use the native python operations on list. I am using Python 3.

Have you attempted a numpy solution yet? If so, which part are you stuck on? This question is a little vague. — Chris Farr
– Chris Farr, Commented Jun 21, 2019 at 11:41
I iterated through the array in the same way. Can't find a vectorized solution. — Akash Kumar
– Akash Kumar, Commented Jun 21, 2019 at 11:42

user10325516 · Accepted Answer · 2019-06-21 12:25:21Z

1

Solution with numpy may look like this:

filtered_array = array[
    (np.abs(array[:, 0, 0] - array[:, 3, 0]) < 5) &
    (np.abs(array[:, 0, 1] - array[:, 3, 1]) < 5), :, :]

where array = np.array(box).

I guess this solution will be significantly faster than plain python if you have prepared data (numpy array). Transforming data from python list to numpy array will nullify any time gain.

answered Jun 21, 2019 at 12:25

user10325516

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

Comments

BramAppel · Accepted Answer · 2019-06-21 11:59:29Z

0

This list comprehension replaces your for-loop construct:

import numpy as np

box = [
      [[4, 4], [4, 8], [8, 8], [8, 4]],
      [[8, 8], [8, 16], [16, 16], [16, 8]],
      [[20,16],[20,20],[24,20],[24,16]]
      ]

box = np.array(box) # convert to numpy array

# one-liner
filtered = [i for i in box if np.ptp(i[:,0]) <= 5 and np.ptp(i[:,1]) <= 5]

filtered = np.array(filtered) # if you want to convert to a numpy array

answered Jun 21, 2019 at 11:59

BramAppel

1,3561 gold badge11 silver badges22 bronze badges

Collectives™ on Stack Overflow

Filter 3D python list using list or numpy

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related