UIAxes 3D scatter plot becomes simplified when rotating with the mouse (App Designer)

Hello,

I am displaying a 3D point cloud in App Designer using a tiledlayout with a UIAxes with MATLAB 2024b. The point cloud is plotted using scatter3.

When the figure is displayed normally, all points appear as in the image below (the volume of interest is circled in red):

However, when I use the mouse rotation tool on the axes, the rendering changes and the plot becomes visually simplified, as illustrated below:

It looks like MATLAB reduces the number of visible points or changes the rendering during the interaction.

Interestingly, the simplified rendering actually looks cleaner and closer to what I want, so I would like to obtain this second rendering as the default display, even when the axes are not being rotated.It is to be noted that I have not noticed this effect when displaying the cloud point in a separate figure.

I would like to know if there is a way to force MATLAB to use this simplified rendering permanently?

Thank you for your help.

Cheers,

Guillaume.

[dpb]

I can't see the simplification of which you speak?

[Guillaume]

The simplification discussed here concerns the 3D volume shown with the jet colormap, circled in red (the other 3D points form a sparse point cloud, so the effect is less noticeable there). This 3D volume represents an isotropic emission, with a gradient in counts from the center to the borders. I probably should have included a zoomed-in view of this area.

Below, you can see the 3D volume with the standard display, where the center is “hidden” by the outer layers, making the highest-density region not visible.

In the next image, when using MATLAB’s rotation tool, the rendering changes: the number of displayed points appears to be reduced, allowing the center of the volume to become visible. This is actually the type of visualization I would like to obtain by default. However, this effect only occurs during mouse-driven rotation. It seems that MATLAB temporarily simplifies the 3D rendering, likely to improve performance during interaction, which results in a clearer view of the internal structure.

Cheers,

Guillaume.

[dpb]

Dunno that anybody here can do anything w/o the dataset to play with...

If doing the rotation programmatically doesn't produce the result desired, I suspect that contacting Mathworks support would be the only way(*) to determine if there is indeed some such reduction applied and to learn what might be being done internally.

However, I suspect if you had another sample dataset, the chances are whatever effects there are in this image may well not appear in another with just a somewhat different distribution of data point locations so I expect you're chasing a mirage.

If you enlarge that (or any other) area can you actually see clearly enough the individual points and discover that not every point really is not being displayed or is it, as I suspect, just an optical illusion? If there really are fewer points being plotted, then you should be able to discern which are missing and discover whether it is something like just a decimation by some factor or if points are being screened perhaps on basis of their Euclidean separation. You could then try to do a programmatic operation similar in nature and see how that works.

(*)I haven't searched to see if perhaps the internal callback of the rotation tool might be m-code or whether it is exposed at all -- programmatically it would seem that perhaps figuring out if there's a way to call that code directly or being able to see what it does would be the only way to reproduce identically what is happening.

[Guillaume]

I can share the dataset if you're interested, but I wasn't sure how relevant it would be.

I've also tried using other datasets and observed the same behavior.

In parallel, I attempted to add listeners to the axes and figure properties, but they didn’t provide any useful clues.

I also tried to reproduce the effect programmatically, but nothing matched the behavior triggered by mouse rotation.

I will contact MathWorks support.

[dpb]

Good luck...will be interested to hear what they have to say.