Add vortex optical flow example

[jni]

Fixes/Closes

Add example of optical flow to the gallery, based on scikit-image.

Epilepsy warning: the movie has some bright/dark flashing.

Screen.Recording.2023-07-06.at.8.05.37.pm.mov

I can remove the reverse phi stuff if people think it's too annoying/dangerous. But it's cool, thought I'd pull up a flashback from my postdoc 😂

[codecov]

Codecov Report

Merging #6041 (262e5a2) into main (72538e5) will decrease coverage by 0.07%.
Report is 1 commits behind head on main.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main    #6041      +/-   ##
==========================================
- Coverage   91.98%   91.92%   -0.07%     
==========================================
  Files         596      596              
  Lines       52638    52638              
==========================================
- Hits        48421    48389      -32     
- Misses       4217     4249      +32     

see 12 files with indirect coverage changes

[kephale]

This is a useful example. The flashing during playback should be opt-in.

[jni]

I was wondering how I would make it opt-in @kephale... Do I just add both the normal contrast frames and the concatenated frames and make the concatenated version invisible? I could even add only the reverse contrast frames with an offset. But I worry both those options will be confusing...

I could also get rid of it altogether... But it's fun to see, no? 😃

[kephale]

I could see something like this helping in the long term: #6045

An short-term alternative is to use a colormap for vortex_phi like twilight that wraps around the color values.

[kephale]

Ok, poked a little more and I'm leaning to:

import numpy as np
from skimage.data import vortex
from skimage.registration import optical_flow_ilk
from scipy.ndimage import zoom

import napari

#######################################################################
# First, we load the vortex image, and then we create an inverted
# "negative" image. For the illusion, we concatenate them end-to-end.

stretch_factor = 5

vortex_im = np.asarray(vortex())
vortex_im = zoom(vortex_im, (stretch_factor, 1, 1), order=1)
vortex_neg = np.max(vortex_im) - vortex_im
vortex_phi = np.concatenate((vortex_im, vortex_neg), axis=0)

The vector lines need tweaking to be visible, but it still has the cool visual effect with less strobing. However, I would still be pretty tempted to not show vortex_neg.

Screen.Recording.2023-07-08.at.8.27.18.AM.mov

[jni]

@kephale

An short-term alternative is to use a colormap for vortex_phi like twilight that wraps around the color values.

Then you would lose the illusion which is the whole point of the thing! 😂 The effect only works when you invert the contrast of the whole image. The only way to make it less stroby is to use a different image that has ~50% white pixels. And I'm not 100% convinced it would work anyway. I'm happy to drop it, I just thought it was a cute application and wanted to write it down somewhere. 😃 I'll do the standard vortex here and save the reverse-phi for my (long neglected) napari-demos repo.

Also what on earth did you do to my vectors 😂

[kephale]

Screen.Recording.2023-07-09.at.8.49.08.AM.mov

This seems to keep a balance between maintaining the illusion and having a gentle transition.

Yeah, vectors are super borked right now.

[edit: code here https://gist.github.com/kephale/8707eb105fc8c52168c1db16ac836580]

[jni]

@kephale @napari/core-devs I've removed the reverse-phi illusion because (a) it was a bit controversial and (b) anyway it made the example more complicated than it needed to be.

The reverse-phi demo lives on at https://github.com/jni/napari-demos/blob/main/vortex.py

Since @kephale already approved this and it doesn't affect the main code in any way, I'm going to add the "ready-to-merge" label.

[Czaki]

@jni could you fix milestone?

[Carreau]

This got the "Ready-to-merge" label 2 days ago and got no objections. Merging to move the queue forward.

Thanks.