Scaling error

[modaresimr]

Description

It seems that scaling parameter is not working correctly. It should be sx,sy,sz but it is sz,sy,sx

What I Did

import numpy as np
import matplotlib.pyplot as plt
from skimage import draw
siz=200

img2 = np.zeros((siz, siz//2,1))
rr, cc = draw.ellipse(siz/2, siz/4, siz*.4, siz*.2, img2.shape[:2])
img2[rr, cc,:] = 1

def myshow(img,spacing):      
        import k3d        
        plot = k3d.plot()
        plot+= k3d.voxels(img, opacity=0.1,scaling=spacing)
        plot.display()
myshow(img2,(1,2,1))
myshow(img2,(100,2,1)) # it should scale x but actually it scales z
myshow(img2,(1,1,2)) # it should scale y but actually it scales x

outputs

[artur-trzesiok]

Hi! Thanks for sharing

myshow(img2,(100,2,1)) # it should scale x but actually it scales z

but second images show scaling in x. Look at:

myshow(img2,(1,1,2)) # it should scale y but actually it scales x

No, it should scale in z axis (1,1,2) -> (x,y,z). And that is exactly what is done:

[modaresimr]

I think there is a mis-interpretation between me and you: I explain my perspective: When we say voxel size it means that each voxel will have a bigger size, for example, if we have a voxel at position (0,0,0) to (1,1,1), for any voxel size, for all axis, in the axis labels we have (1,1,1) but from the visualization point of view, we will have a bigger voxel.

Therefore, First: In all of the above cases, the axis labels should be the same since we only change the voxel size

second: let's see img2(1,2,1):

If you compare case myshow(img2,(1,2,1)) and myshow(img2,(1,1,2)), you can see that the only difference is that scale down y and scale up z.

what is your expectation? it should have more depth (z), and it should only scale down y by 2 and x with no change:

[artur-trzesiok]

@modaresimr I'm not sure what exactly do you want to get as outcome :(. Can you share some example data or sketch in paint?

Please make sure that we understand two topic similary:

First:

in k3d z axis is directed to "up"

Second: In k3d when we talk about numpy dense data we have convention that shape is mapped as (L, H, W). So

img2.shape
(200, 100, 1)
z_size, y_size, x_size = img2.shape

That's exactly why our ellipse is displaying on YZ plane

[artur-trzesiok]

@modaresimr did that make sense to you?