Now that all the IRTC submissions are in, I can reveal my technique for
getting millions of POV primatives into a render without excessive amounts
of memory to parse and render the image.  I was so excited with the
technique and results that I built a web-site to describe it.

http://www.angelfire.com/md2/castlewrks/

I encourage everyone to try them out and hope to see more detail in
everyone's pictures.

Comments as always are welcome!

Robert J Becraft
aka cas### [at] aolcom

Post a reply to this message

Robert,
Your images are impressive and beautiful. The technique is brilliant.
My one observation is that it might prove problematic in stereoscopic
rendering. The actual shape of the object, flat, square, etc. will be
more apparent when viewed in stereo. Maybe you could do a test. Render a
second version of the garden image and shift the camera a little to the
left, then view it stereoscopically.

Harold

Robert J Becraft <cas### [at] aolcom> wrote in message
news:38710420@news.povray.org...
> Now that all the IRTC submissions are in, I can reveal my technique for
> getting millions of POV primatives into a render without excessive amounts
> of memory to parse and render the image.  I was so excited with the
> technique and results that I built a web-site to describe it.
>
> http://www.angelfire.com/md2/castlewrks/
>
> I encourage everyone to try them out and hope to see more detail in
> everyone's pictures.
>
> Comments as always are welcome!
>
> Robert J Becraft
> aka cas### [at] aolcom
>
>
>

Post a reply to this message

"Harold Baize" <bai### [at] itsaucsfedu> wrote:
>
> My one observation is that it might prove problematic in stereoscopic
> rendering. The actual shape of the object, flat, square, etc. will be
> more apparent when viewed in stereo.

Or in many cases if there are reflective surfaces involved.


-- 
Jeff Lee         shi### [at] gatenet         http://www.gate.net/~shipbrk/

Post a reply to this message

> Or in many cases if there are reflective surfaces involved.

What about shadows? They'd be different too. It is still a good method
though, don't get me wrong.

--
Homepage: http://www.faricy.net/~davidf/
___     ______________________________
 | \     |_       <dav### [at] faricynet>
 |_/avid |ontaine      <ICQ 55354965>

Post a reply to this message

Robert J Becraft wrote:
> 
> Now that all the IRTC submissions are in, I can reveal my technique
> for getting millions of POV primatives into a render without excessive
> amounts of memory to parse and render the image.  I was so excited
> with the technique and results that I built a web-site to describe it.

These are good techniques, but I have a minor nit to pick.  What you're
really doing is putting the *appearance* of millions of primitives
into the scene, by painting an image map onto a polygon, and placing it
appropriately.  More precisely, you're making one primitive do the
work of hundreds, and are arranging things so that the qualitative
difference is not easily perceptible.

I did a similar thing in three of my IRTC animations.  I modeled an
orchard of petrified trees, using a single mesh of smoothed triangles.
They are fully three-dimensional (which is needful for an animation).
There are three different ones, and they are rotated and scaled to
disguise their similarity.  The vertex data is only stored for the
first copy of each of the three trees; the other instances only have
texturing and transformation data attached to them.  I didn't put leaves
on them (they were petrified), but that would not have caused memory
use to rise all that much.

I also did the same thing for the Ant people a couple rounds back;
after modeling each part of the bug with bicubic patches, I converted
them to meshes (using a macro), and then simply copied the parts over,
texturing them differently from one creature to the next; using
bicubic patches for every part of each bug would have consumed
4.5 megs of memory for each bug.  Using meshes required about 3.5
megs for the first bug (since pieces on the left and right sides of the
bugs are mirror images of each other, they could be re-used within
each bug), and about 50K for each bug after that.

The tractor tread pieces on Rusty were first a CSG union of six boxes,
nine cylinders, and four spheres, for nineteen objects altogether.
There are 56 of these pieces in Rusty, and these 1064 objects were
taking up memory on my 16 Meg 486 system; I converted the same union to
a mesh of smoothed triangles, and got a vast savings in memory.

The memory savings available when making copies of meshes is one of the
chief advantages of using them.  I don't have a mesh editor, so the
meshes had to be generated through scene code, but the nature of the
objects to be modeled made this fairly simple.

Regards,
John

Post a reply to this message

David Fontaine <dav### [at] faricynet> wrote:
>
>> Or in many cases if there are reflective surfaces involved.
>
> What about shadows? They'd be different too. It is still a good method
> though, don't get me wrong.

Oh, certainly, I agree.  Like most things, it's just better suited for
certain applications, and not quite as well suited for others.


-- 
Jeff Lee         shi### [at] gatenet         http://www.gate.net/~shipbrk/

Post a reply to this message

>>These are good techniques, but I have a minor nit to pick.  What you're
>>really doing is putting the *appearance* of millions of primitives
>>into the scene, by painting an image map onto a polygon, and placing it
>>appropriately.  More precisely, you're making one primitive do the
>>work of hundreds, and are arranging things so that the qualitative
>>difference is not easily perceptible.

The image being used is composed of several thousand objects that I would
have placed in the actual composition as primatives IF POV could handle that
many objects.  Since there was no way to do that, in the composition I was
creating, I experimented and discovered a technique that allows me to give
the appearance as you say of millions of primatives.

It is a technique and because it is a technique, it may not apply to
everyone's needs.  Where it doesn't work, for example in an animation, it
will either need to be adapted or not used.

The main and intended purpose of this technique is to allow organic
materials which by their very nature are primative object intensive to be
modelled and incorporated into scenes that require that high level of object
primatives.

People can now cover their castles with ivy.  You can have trees in the
distance of a previously barren landscape.  You can have more than 1 tree in
your barren landscape.   For my purposes, I have millions of flowers in my
garden.

Robert J Becraft
aka cas### [at] aolcom

Post a reply to this message