"Dre" <and### [at] gmailcom> wrote:
> Really, they are WOW images to me :)

Thanks! It's just a shame they don't look as good at higher resolutions.

> I'd love to know some more about how you did this!

Most of the work is done externally, I'm really only using POV-Ray to plot a
dataset and do pretty lighting. I used some of the equations given on the
Chaoscope website (http://www.chaoscope.org/doc/attractors.htm), originally in a
Java app (the quickest way I knew to implement some sort of gui at the time).
Basically, if the orbit doesn't collapse or escape within a few hundred
iterations I plot it properly for inspection. Most of these turn out to be
mundane, but every now and again an 'interesting' one crops up. At this point I
save the equation coefficients in a data file.

Java is entirely unsuitable for iterating 1e8 points over a 30+ coefficient
dynamical system, so eventually I wrote a simple tool in C++ to do some proper
plotting from the saved coefficients, and (optionally) to write a text file full
of vectors suitable for reading from POV-Ray SDL.

Add some random lights using the data file as a seed, and away we go...

:)

Post a reply to this message

Tim Cook <z99### [at] gmailcom> wrote:
> Now model a modern art gallery and put these on pedestals therein.

Good idea! I could make some of them really big, put 2D plots in picture
frames... *gazes into middle distance*

Post a reply to this message

Christian Froeschlin <chr### [at] chrfrde> wrote:
> Neat! I take it these images are essentially an exorbitant
> number of spheres around externally generated data points?

Exactly right. :) (See my other post for my somewhat hacky workflow)

The data files vary in size from a few 100k for the thin spindly ones, up to
20MB+ for the more voluminous ones. Luckily even the vast ones are pretty quick
to produce from just the equation coefficient data. The slowest part appears to
be writing the data file!

> I wondered if it's a blob but I thought some transitions
> didn't quite look like it, although it's hard to tell
> with the small object size. If it's spheres it might
> be worth to try a blob for faster rendering.

I tried blobs, but they turn out to be many times slower for this case. These do
use radiosity and multiple area_lights, after all. With antialiasing off, they
render in a couple of minutes (parsing can be longer!). With AA on, an hour or
so.

Actually I quite like the expanded-polystyrene/gritty look that bare spheres
produce!

Post a reply to this message

I Really love "barbwire"... How did you code these ?


> All that amazing mandelbulb stuff has inspired me to dig out a strange 
> attractor search/plot app I wrote a few years ago, and try to get some 
> decent plots into POV-Ray. Here's a few of the nicer ones I've found 
> over the last couple of weeks.
> 
> Bill
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
> 
> 
> ------------------------------------------------------------------------
>

Post a reply to this message

Bill Pragnell <bil### [at] hotmailcom> wrote:
> All that amazing mandelbulb stuff has inspired me to dig out a strange
> attractor search/plot app I wrote a few years ago, and try to get some
> decent plots into POV-Ray. Here's a few of the nicer ones I've found
> over the last couple of weeks.
>
> Bill

Excellent really interesting shapes, I don't think I will even try to understand
how they were made and just accept I should have paid more attention at school.
It's a shame they don't look as good at high resolutions you could make some
fairly good abstract poster art.

Sean

Post a reply to this message

> Tim Cook <z99### [at] gmailcom> wrote:
>> Now model a modern art gallery and put these on pedestals therein.
> 
> Good idea! I could make some of them really big, put 2D plots in picture
> frames... *gazes into middle distance*
> 
> 
Beter yet:

Make them *for real!*


Alain

Post a reply to this message

> Christian Froeschlin <chr### [at] chrfrde> wrote:
>> Neat! I take it these images are essentially an exorbitant
>> number of spheres around externally generated data points?
> 
> Exactly right. :) (See my other post for my somewhat hacky workflow)
> 
> The data files vary in size from a few 100k for the thin spindly ones, up to
> 20MB+ for the more voluminous ones. Luckily even the vast ones are pretty quick
> to produce from just the equation coefficient data. The slowest part appears to
> be writing the data file!
> 
>> I wondered if it's a blob but I thought some transitions
>> didn't quite look like it, although it's hard to tell
>> with the small object size. If it's spheres it might
>> be worth to try a blob for faster rendering.
> 
> I tried blobs, but they turn out to be many times slower for this case. These do
> use radiosity and multiple area_lights, after all. With antialiasing off, they
> render in a couple of minutes (parsing can be longer!). With AA on, an hour or
> so.
> 
> Actually I quite like the expanded-polystyrene/gritty look that bare spheres
> produce!
> 
> 

You could output as DF3 file and render as an isosurface. You also could 
render those as media. It can be worth a try.



Alain

Post a reply to this message

Alain <aze### [at] qwertyorg> wrote:
> You could output as DF3 file and render as an isosurface. You also could
> render those as media. It can be worth a try.

That was the first method I tried. However, the df3 is very wasteful since most
of the volume is empty. Furthermore, the df3s for these examples would be at
least 512MB each! During testing at ~100x100x100, I found that generating,
reading in and rendering a df3 was much slower than the method I'm currently
using for a volume 64x larger.

I'm using a 3d vector<bool> to store the volume during iteration, then only
write out the voxel coordinates that were hit. That way, memory usage and file
size is at a minimum and I throw away all the empty locations. Unfortunately, I
also throw away the cumulative information, but it's very quick to iterate so
I'll probably come back and do something with it in future.

Bill

Post a reply to this message

LightBeam <seb### [at] gmailcom> wrote:
> I Really love "barbwire"...

Yes, I like that one too - looks like the Sorting Hat from Harry Potter!

>How did you code these ?

The equations can be found here:

http://www.chaoscope.org/doc/attractors.htm

Using different coefficients in these equations produces different attractors
(although most combinations don't produce an attractor - you need to 'search'
random combinations of coefficients to find interesting shapes). Each new point
is calculated from the previous point, and the attractor is built up point by
point. It's just a matter of storing those points and plotting them, either as a
2D projection, or in 3D.

I wrote a java app some time ago to search for and plot attractors, but I had to
rewrite the plotting in C++ (for speed) to get data files that I could then read
into POV-Ray.

Bill

Post a reply to this message

Looks like a job for a 3D printer!   :-)

Here's one for only $5000:  http://www.desktopfactory.com/our_product/



On 2009-12-07 21:14, Alain wrote:

>> Tim Cook <z99### [at] gmailcom> wrote:
>>> Now model a modern art gallery and put these on pedestals therein.
>>
>> Good idea! I could make some of them really big, put 2D plots in picture
>> frames... *gazes into middle distance*
>>
>>
> Beter yet:
>
> Make them *for real!*
>
>
> Alain

Post a reply to this message