How to resize 4D-image using numpy?

Question

I have (5000, 32, 32, 3) as numpy.ndarray. (number of images:5000, image size: (32x32), channel: 3)

I want to resize the image (32x32) to (28x28). How can I do this?

This is way too broad a question for a good answer, but I can point the way here and here. And for the future, please read How to Ask and minimal reproducible example — Daniel F
– Daniel F, Commented Sep 24, 2018 at 6:22

norok2 · Accepted Answer · 2018-09-24 07:25:58Z

1

Something like:

import scipy.ndimage

factor = 28.0 / 32.0
scipy.ndimage.zoom(input_array, (1, factor, factor, 1))

answered Sep 24, 2018 at 7:25

norok2

27.1k6 gold badges83 silver badges110 bronze badges

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

2 Comments

Do note that zoom is "lazy" when downsampling, not properly averaging over pixels and hence losing some information.

Sure, running a uniform (or other) filter prior to to zoom would indeed be a good idea.