whXsiIce_NeuralNetwork BETA

User avatar
whurst
Posts: 43
Joined: 16 Dec 2012, 11:00
Location: Germany

whXsiIce_NeuralNetwork BETA

Post by whurst » 05 Oct 2016, 20:35

Hi guys,

i made a thing :-)

a small one.

I do not have any tutorial videos,
i do not have any high res awesome pictures,
i do not have any super cool documentation,

but i have a ICE (finally) node,
well sure it's not big
and it's not complicated,
and sure it has absolute nothing production related

but finally it loads and stuffing around :-)

OK ... let's get serious ...

I wrote a small plugin that generates points on an imaginary "edge" between all input points.
I did this with ICE and it takes ages - so i write this plugin to get more power.
The particle size is change in depend of the distance to another particle.

Result is some kind of this (using emPolygonizer ofcause - Thanks Eric)

EDIT: 08.07.2017 - Image removed - Please wait a moment for the release

It is BETA and i do not know if someone need this, or try this, or ever use it.
I made a small overview page of this plugin where you can download it
and watch some strange screenshots i made

EDIT: 08.07.2017 - URL removed - Please wait a moment for the release

XSI 2012 32/64bit

Have Fun :-)

EDIT: 08.07.2017 - New website did not contain the BETA anymore. But i have a new plugin in my queue (somewhere) that contains neural network ... So please wait ... sorry ...
Last edited by whurst on 08 Jul 2017, 10:42, edited 1 time in total.

User avatar
mc_axe
Posts: 405
Joined: 12 Mar 2013, 18:44

Re: whXsiIce_NeuralNetwork BETA

Post by mc_axe » 06 Oct 2016, 00:46

I like abstract organic stuff always
I alrdy imagined some scene with uhmm organic space frames ,,
i dont know exactly what i mean by that,,, i imagined something artistic
But google tells me ppl use those in architecture lol
Image

You can also make low poly versions of something lets say Eiffel tower with maybe 250 points and model/render/3d print a more organic version.
Last edited by mc_axe on 07 Oct 2016, 19:09, edited 1 time in total.

User avatar
gaboraa
Posts: 272
Joined: 16 Apr 2010, 23:14

Re: whXsiIce_NeuralNetwork BETA

Post by gaboraa » 06 Oct 2016, 14:05

Cool stuff, thanks a lot for sharing!
You do not have the required permissions to view the files attached to this post.

User avatar
Maximus
Posts: 1006
Joined: 09 Jun 2009, 15:45

Re: whXsiIce_NeuralNetwork BETA

Post by Maximus » 07 Oct 2016, 02:03

thanks a lot for sharing :)
altho it doesnt work on si2013, i guess its not supported but great work nonetheless!

User avatar
rray
Moderator
Posts: 1684
Joined: 26 Sep 2009, 15:51
Location: Bonn, Germany

Re: whXsiIce_NeuralNetwork BETA

Post by rray » 07 Oct 2016, 02:04

Thanks, looking good.
softimage resources section updated Feb 7 2019

User avatar
mc_axe
Posts: 405
Joined: 12 Mar 2013, 18:44

Re: whXsiIce_NeuralNetwork BETA

Post by mc_axe » 07 Oct 2016, 19:00

Hello iv tried the demo scene on 2015sp1 with the 64bit 2012 dll i get insta crash when i unhide the "test_06_create_particle_cloud_without_cache" model

A few questions what the 750 stands for in dlls?
Why in some examples the particles follow all vertices on input geometry like the "test_04_turbmesh_as_basepoints_with_animation" , and in others i see only a few points even if input is a sphere with many more vertices?

And to add 2 cents on wish list:

I think atm you get like all possible combinations between vertices so its hard to get the outline of a character for example.You could make another distance threshold. Something like generate particles "on imaginary edge" only if distance between input points < X

User avatar
whurst
Posts: 43
Joined: 16 Dec 2012, 11:00
Location: Germany

Re: whXsiIce_NeuralNetwork BETA

Post by whurst » 07 Oct 2016, 19:27

Hi mc_axe,

Well 2015sp1. I didnt have the SDK files, so i could not compile this for 2015. That can maybe the problem.
Test 6 contain emPolygonizer setup ... maybe this is incompatible stuff that works in 2012 but not longer in 2015 :-()

The 750 stand for XSI 7.5 and it is for XSI 7.5 and Mod Tool 7.5. But i will discontinue these Versions, because there are some very important stuff missing.

In all demos (except this part 4) i generate just a few points on the surface of the sphere ... i do not use the points of the sphere - well this would break the PC if i would do this :-)

In some parts i use generate samples in other i use a separate point cloud with an emit in it an import this point cloud.

Your Wish:

something like this ?

If "virtual edge length" less "5" than "skip generate particles"

? that sounds easy :-) please confirm and i will do it that way
mc_axe wrote:Hello iv tried the demo scene on 2015sp1 with the 64bit 2012 dll i get insta crash when i unhide the "test_06_create_particle_cloud_without_cache" model

A few questions what the 750 stands for in dlls?
Why in some examples the particles follow all vertices on input geometry like the "test_04_turbmesh_as_basepoints_with_animation" , and in others i see only a few points even if input is a sphere with many more vertices?

And to add 2 cents on wish list:

I think atm you get like all possible combinations between vertices so its hard to get the outline of a character for example.You could make another distance threshold. Something like generate particles "on imaginary edge" only if distance between input points < X

User avatar
mc_axe
Posts: 405
Joined: 12 Mar 2013, 18:44

Re: whXsiIce_NeuralNetwork BETA

Post by mc_axe » 07 Oct 2016, 19:31

If "virtual edge length" less "5" than "skip generate particles"

? that sounds easy :-) please confirm and i will do it that way
Yes exactly! :-bd

User avatar
whurst
Posts: 43
Joined: 16 Dec 2012, 11:00
Location: Germany

Re: whXsiIce_NeuralNetwork BETA

Post by whurst » 07 Oct 2016, 20:20

mc_axe wrote:
If "virtual edge length" less "5" than "skip generate particles"

? that sounds easy :-) please confirm and i will do it that way
Yes exactly! :-bd
OK :-) Here you go

Just replace the DLL with the new one and you can setup this.

New test model is available and new screenshot

edit: 20.02.2018 - missing png removed, because reorg of website
Last edited by whurst on 20 Feb 2018, 20:02, edited 1 time in total.

User avatar
mc_axe
Posts: 405
Joined: 12 Mar 2013, 18:44

Re: whXsiIce_NeuralNetwork BETA

Post by mc_axe » 07 Oct 2016, 23:35

That was quick! Nice addition it gives alot of versatility in the end shape, rly good for spaceframes :D
And also good to keep the complexity at the surface if you want to keep the inside of a volume empty.

Here is same input null points, i play only with the min&max edge length filters
Image