Ready - The return of mcjIsosurface plugin for PC-Win and DS 4.X

mCasual

jag11 said:

Awesome! Video at 0:16 seems like a heart beating. Please post the script to do that.

i called it the boiling surface effect

 it's metaballs bouncing up and down

one of the problems with those scripts is that i was adjusting them "by hand" and you need to be a daz script programmer to do that

but the plan was to make them usable by anyone and maybe have 1 script for all effects

if it's not too much work, i'll post the boilingplate effect script,

but it may have to wait untill i complete the task at hand:  the mcjIsoSurface for IRay update

 

mCasual

i found the (2011) script for the boiling-plate effect and it works well with the mcjIsosurface4

it's not adjustable but someday will be

 

see if your daz studio 4.8 Preferences/Interface/Pixel Shading" option is ON and the Reflection Map of a surface has an image applied to it, you get Real-Time Display of the chrome/mirror effect

though now if i remember well you get the same effect by applying it to the diffuse channel --- because the isosurface uses a spherical UV map

mCasual

first animated blob ( it's just paint !!! ) rendered with IRay

 

oh now i understand why the metaballs are not jumping !

i prevented them from reaching the floor !

contrary to initial intent, i'll release the script for this effect with a User Interface

so you'll be able to adjust things

 

mCasual

it's more complicated than just running the script,

but, you get to tailor the effect to your liking

the script will be posted on a different page than the isosurface plugins ( because they have other uses! )

here we'll make a 30 frames long effect

checking the metaballs paths through a 30 frames "cycle" and we can see it's not looping

Since it uses parabolas and random initial speed and directions, it's understandable that we dont get bounces with very predictable durations

but but but i could weed out the bounces that are not fully encompassed by the looping animation-range

 

so instead of Y.O.L.O. it would be YOLAINOTWTAR You Only Live An Integral Number Of Times Within The Animation Range

 

But ah HA! back in 2011 i wrote the looping version!

mCasual

n

mCasual

mcjFXBoilerPlateB

The first mcjFX script which creates animated metaballs ... so that the Isosurface looks like this

was posted a moment ago !

https://sites.google.com/site/mcasualsdazscripts6/mcjfx

 

there may even be a manual tomorrow

in the meantime you'll have to figure the settings or leave them un-modified ( and there's a reset button ... which i didnt test actually )

 

jag11

Thanks for the mcjFX.

mCasual

someday someday i'll rewrite the polygonizer part of the plugin and get rid of the mataAnnoyingLines

in other isosurface news, - next version will have the visible domain box and sane transforms handling

mCasual

but but but

as we can see here upon a close inspection

is that the lines are not only due to the variable spacing between edges

there's vertex positioning imperfections ... possibly due to a loss of precision somewhere ..

maybe if i fix that, the appearance will be acceptable!

 

mCasual

this weekend i'll work on the more-iRay-Friendly mcjIsosurface4

since i didnt remove the code for it, the mcjIsosurface node still renders itself in 3Delight mode

the top sphere is the mcjIsosurface node, as you can see, the plugin does a better job, because it uses computed normals instead of facet normals

the bottom sphere is the "it" node which was hijacked by mcjIsosurface and is rendered by Daz Studio and/or iRay, like any normal primitive

--

changes i expect to do this weekend

- make the mcjIsosurface and "it" node positions, rotations and scaling conform to one another

- make sure it's possible in 3Delight mode to turn on/off the mcjIsoSurface old but good rendering

- fix the imprecision in vertex positioning

- possibly add a decimation option where the plugin merges all the vertices that have a distance smaller than a certain distance

 

i did a test, i created an isosurface sphere

exported the isosurface using Daz Studio's obj exporter

imported it back in Daz Studio

ran a script that measures the radius at each vertex

and i can report that the radii were

Executing Script...min max = 9.995745658874512 10.013422966003418

that's in centimeters

 

 

 

mCasual

i'm not there yet, but if all goes well the result will be like the ball on the right

instead of the one on the left

but you could say this was fixed "manually" i still didnt find what's wrong with mcjIsosurface3 and 4 

maybe it's something that appeared in DS4.5 or 4.7 ... tests to do

mCasual

this post was a duplicate

hphoenix

The ball on the left looks like the marching-cubes aren't smoothing properly....the one on the right has some normal issues, but appears smooth.

 

mCasual

note that this was using the "metanuball" 

i tested mcjisosurface3 in DS3-64 bit and it's the same

the "metaball" primitives seem to look similar to the ball on the right

by stepping up the resolution of the mcjIsosurface ( in the parameters tab ) the defects become less apparent

----

dont know yet if there will be anything i can do to improve the looks other than changing the polygonizer

note that my reference for all this is http://paulbourke.net/geometry/polygonise/

----

one thing i may do is reduce the number of edges by not triangulating the quads

this should smoothen the meshes a bit

http://mathematica.stackexchange.com/questions/80669/integrating-a-function-over-a-surface-integral

 

 

 

 

hphoenix said:

The ball on the left looks like the marching-cubes aren't smoothing properly....the one on the right has some normal issues, but appears smooth.

 

 

mCasual

laboratory "monster" while trying the repair the metanuball primitive

 

today's conclusion is that i'm not certain what i was trying to accomplish with the metanuball primitive

but until i do, you can get good results using the metaball and metasphere primitives

i probably left the metanuball in the mcjMetaConsole palette for backward compatibility

--

tomorrow i'll study other ways to improve mcjIsosurface

i did a few modifications on my copy that do help, and at a minimum you'll get those tomorrow

or if you're really really insisting, here's the nightly build for 64-bit

mcjIsosurface4 64 bitearly 08292015 Morning zip 

 

mCasual

the laboratory "monster" has some cuteness , mathematically speaking i mean, no quasi-degenerate thin-sliver-triangles,

 

hphoenix

Hmm.  And have you considered an octree traversal algorithm instead of marching cubes?  There's plenty of info on variations used for isosurface reconstruction, and can easily result in considerable speedups from the brute-force marching cubes.  As an example:  https://users.soe.ucsc.edu/~avg/Papers/oct92.pdf

 

RAMWolff

Interesting project.  Cool too! 

mCasual

i think i'm using marching tetrahedrons, but i dont want to spend more than this weekend on the isoSurface plugin

i actually have projects on the backburner goiung back to oh about 2010 .... that's a lot of baccklog to tend to

 

my mcjCollider plugin and some scripts would benefit from speedups offered by spatial partitioning

 

it's all free/hobby/toy gizmos so i'm happy with them as they are

 

 

 

hphoenix said:

 

Hmm.  And have you considered an octree traversal algorithm instead of marching cubes?  There's plenty of info on variations used for isosurface reconstruction, and can easily result in considerable speedups from the brute-force marching cubes.  As an example:  https://users.soe.ucsc.edu/~avg/Papers/oct92.pdf

 

 

mCasual

good news, our research team ( thinks it ) has located the source of the isosurface4 (and 3 i guess) banding defects
there was a line in the code that looked like d += 0.0001;
i was adding a small tweak to avoid zero division or some "fence" issues
but apparently 0.0001 is not small enough of a tweak to go unnoticed visually
so i changed it to d += 0.000001 and looks like that does the trick
the updated version will be posted in the daz3d freepository forum later today
i want to improve the domain vs isosurface  relations
iRay render here
the red ball is the isosurface, since it exceeds the "domain" it gets truncated
the Utah teapot ( 4 pieces ) is a free prop on my web site

mCasual

for a moment there i had doubts !

the defects were so small

but in this extreme closeup of a small metaball

you see in red an mcjIsosurface4 ( with the corrected code )

and in white an mcjIsoSurface3 ( with the old code )

and there is a difference

so we indeed fixed that issue

 

fig 2 - the 'metaball' primitive is one of the few primitives which given a negative radius give us a subtractive effect

 

fig 3 the red metaball on the left is a "metanuball", the one on the right is a standard metaball

the banding on the metanuball is not due to a problem with the isosurface,

the metanuball was an experimental thing and in fact i dont quite remember what was the goal

 

RAMWolff

I have no idea about any of this but as usual your brilliant and thanks  to your team for their discovery!

mCasual

well actually the metacylinder still exhibit the problem ... the case continues

hphoenix

I think the original vertex interpolation code you are using was based on using float and double type elements, back on old hardware.  And it was based around a domain that was fairly large compared to the isolation value (0.0001) to avoid NaN issues.  The double type values were used to give 'room' for division/multiplication results.

 

However, either way, with modern hardware (32-bit/64-bit) a double is a LOT bigger for most compilers.  Instead of using a numeric 0.0001 (or 0.000001), #define a constant in your header files like 

#define MINIMUM_DIV 0.000001

so that you can just change it there, and have all those comparisons use the identifier instead of the direct value.  And since a modern 64-bit double (IEEE 754 standard) can go as low as

0.0000000000000002

in representing fractional parts, you could use a much smaller value (though give a few orders of magnitude above that limit, so that if you scale up the domain, it doesn't become an issue).

Now, depending on which C++ compiler you are using, the "long double" is an option (on just about everything but MS Visual Studio) as it extends the double to a 80-bit field.  In MSVC, a "long double" is the same size as a "double".

 

I'm still working on my algorithm for dynamic octree spatial partitioning so that it works with continuous functions (as most implementations are designed to work with point sets, so it's just a lookup for containment, rather than evaluation a scalar field, which seriously complicates determining if a octant should be subdivided or if it should be considered 'empty'.....) but when I get it worked out, I'll pass you a copy of the sample implementation.  Might be a good optimization for a few of your plugins!

 

Keep up the good work!

 

(and the metacylinder image.....is that TWO metacylinders meeting, or what?  That banding looks to be localized to the extreme curvature region.....)

 

mCasual

it was a birdeye view of a metatube showing a dent

the mataballs now seem to look good

but other metashapes like the metatubes still have the problem

note that when you increase the resolution, or animate the iso or add a texture the problem becomes less visible

i'll give it another go in the next days

 

about the precision, i think using the small number like 0.0000001 was helpful, but in the end i changed the method and dont need it anymore

i did as ... probably Paul Bourke ... did here around line 00508

 

http://www.ks.uiuc.edu/Research/vmd/doxygen/Isosurface_8C-source.html

 

 

 

 

 

 

 

hphoenix said:

 

(and the metacylinder image.....is that TWO metacylinders meeting, or what?  That banding looks to be localized to the extreme curvature region.....)

 

 

mCasual

well ... i seem to have improved the softness of the mataballs primitives but other metashapes like the metaspheres still suffer from ripple/bands defects

tomorrow or next weekend i'll give it another go

i may post the slightly improved plugin later this evening

https://sites.google.com/site/mcasualsdazscripts5/mcjisosurface4-beta-20150802-plugin-for-ds4-8

the current development version 64 bit plugin in a zip file 

 

 

Mythmaker

I just tried the isosurface4 all 3 versions, only the latest beta 03/09/15 version will have the wireframe show up in the viewport.

But wireframe yes/no options didn't show up under Parameters, just Geometry: 'resolution x32' and 'show doman box' only.

I've set default wiremap Reflection map to 100%, diffuse/ambient 0. Metaconsole metaball/cubes etc still don't show in viewport or actual render.

Changing the engine to 3DL did nothing. Changing shader to iray uber base, no change. 

Same with boilerplate. Both the isosurface object and boilerplate do exist in the scene as shown.

Most likely a 4.9 issue....?

In any case, no hurry. Just thought I put the report here for 4.9.2.7.

This is my first attempt at isosurface anyway so let me know if I missed specific steps... Cheers

 

mCasual

looks like it's broken for earlier versions of Ds4.9 too

worked last summer but ...

i suspect the isosurface doesnt "see" the meta-objects 

i think the isosurface plugin scans the scene for objects with specific names like "metaball"

i'f i'm lucky it's just a matter of renaming them ( which is different from re-labeling them )

i'll check this after supper which is now !
 

 

 

Mythmaker said:

I just tried the isosurface4 all 3 versions, only the latest beta 03/09/15 version will have the wireframe show up in the viewport.

But wireframe yes/no options didn't show up under Parameters, just Geometry: 'resolution x32' and 'show doman box' only.

I've set default wiremap Reflection map to 100%, diffuse/ambient 0. Metaconsole metaball/cubes etc still don't show in viewport or actual render.

Changing the engine to 3DL did nothing. Changing shader to iray uber base, no change. 

Same with boilerplate. Both the isosurface object and boilerplate do exist in the scene as shown.

Most likely a 4.9 issue....?

In any case, no hurry. Just thought I put the report here for 4.9.2.7.

This is my first attempt at isosurface anyway so let me know if I missed specific steps... Cheers

 

 

mCasual

end of 2015 following a hardware disaster i lost the DS4.8 / 2015 version source code for this plugin

i still have older DS3 versions and can probably remember how i made it DS4.8 then iRay compatible

but it will take some time before i have time to work on this

 

mtl1

@mCasual, have you considered hosting your code on an online repository to prevent such hardware disasters? :) Aside from your own website, of course!