Help! Ellipses seamless pattern

@laurent_delrieu your definition could work too + @jopsa2 gradual deformation.

For images should be a real ellipse packing, that the ellipse direction could respond and align to some RGB value or something and 4 radii(min-max length and size) in order to have ellipses as circles too.
Hi, @DanielPiker could this be implemented in a custom goal a kind of LineCollider?

Hi @ThomasE

Yes, I can look at adding an ellipse collider with inputs for a direction and anisotropy field as well as size.
For this project we did something similar, but for 3 dimensional ellipsoids.
Essentially it does a scaled sphere packing, like what’s shown in Laurent’s post above, but it does this scaling in way that varies locally, with the axis and factors dependent on the input field.
For 2d ellipses it’s quite a bit simpler than the full 3d case, so I think the code could be simplified considerably.

This would be only for ellipses which are locally similar in orientation and anisotropy like on the left - For packing arbitrary ellipses like on the right you’d need to use the CurveCollider.

thank you for your response @DanielPiker, maybe a good approximation would be if we work with lines and collider.
(length and direction could be controlled with other existing goals),

but I was thinking if you could just add a colour mesh and radius MAX input to your objects collider. Basically, a kind of Image ImageObjectsPacking. I think this could work.

Yes, an image version of the collider with Line segments could be a good addition.
I guess you could also use a diamond to get a closer approximation.

For moderate numbers of ellipses it looks like the CurveCollide isn’t too bad for speed.
It’s much more precise but probably too heavy to use this approach for thousands of them though. (10.9 KB)


:raised_hands:t2:Thank you for considering this.

I tried the diamond approximation with long and short ellipses some as circles but finding the correct pipe radius is madness :sweat_smile:

For this purpose, I think a MIN MAX line radius would be enough. So again the idea that a coloured mesh informs the radius of the line would be really helpful to approximate ellipses and circles.

this is amazing.
Is what im looking for but with ellipses instead.

The idea would be to make something like this but with ellipses tangent inside the boundary instead.
circle_sizegradient (25.3 KB)