replace row of 3D array if all values below test - one numpy line using indexing

Question

Is there a numpythonic way to do this? Original array is of dimension N x M x 3, I'd like to make a N x M mask as shown. I thought there was a simple way to write this with Numpy but I'm having brain freeze.

While there may be alternative ways to implement this, I'm trying to do it this way to maintain my skills at using Numpy indexing.

This fails because img2[test] looses the original shape. Do I have to .reshape() somehow, or is there simpler way where I don't need to explicitly restate the dimensions of the axes?

import numpy as np
import matplotlib.pyplot as plt

img  = np.random.random(100*100*3).reshape(100, 100, -1)
img2 = img.copy()

target = np.array([0.5, 0.3, 0.7])

test = np.abs(img - target) < 0.5

img2[test] = np.zeros(3)  # probem, img2[test] looses its shape

plt.figure()
plt.imshow(np.hstack((img, img2)))
plt.show()

File "/Users/yournamehere/Documents/fishing/Eu;er diagram/stackexchange question v00.py", line 12, in <module>
    img2[test] = target
ValueError: NumPy boolean array indexing assignment cannot assign 3 input values to the 11918 output values where the mask is true

Paul Panzer · Accepted Answer · 2017-02-10 04:42:52Z

1

Here's a one-liner showcasing the very useful keepdims kwarg

np.where(np.logical_and.reduce(test, axis=-1, keepdims=True), [0, 0, 0], img)

answered Feb 10, 2017 at 4:42

Paul Panzer

53.3k3 gold badges60 silver badges103 bronze badges

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

2 Comments

uhoh Over a year ago

Thank you for the mini numpyphany! It isn't every day one can learn several new things from a one-line answer.

Paul Panzer Over a year ago

@uhoh Thanks! Glad you're enjoying it.

uhoh · Accepted Answer · 2017-02-10 02:15:32Z

0

Brain has un-frozen. Forgot to use .all() to collape the test to a single boolean value for each N x M location. Once that is corrected, Numpy does it's thing very nicely.

import numpy as np
import matplotlib.pyplot as plt

img  = np.random.random(100*100*3).reshape(100, 100, -1)
img2 = img.copy()

target = np.array([0.5, 0.3, 0.7])

test = (np.abs(img - target) < 0.5).all(axis=-1)

img2[test] = np.zeros(3)  # probem, img2[test] looses its shape

plt.figure()
plt.imshow(np.hstack((img, img2)))
plt.show()

answered Feb 10, 2017 at 2:15

uhoh

3,8358 gold badges54 silver badges112 bronze badges

Collectives™ on Stack Overflow

replace row of 3D array if all values below test - one numpy line using indexing

2 Answers 2

2 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related