TITLE: warm_up
NAME: Norbert Kern
COUNTRY: Germany
EMAIL: norbert-werner.kern@t-online.de
WEBPAGE: not yet
TOPIC: Insects and Spiders
COPYRIGHT: I SUBMIT TO THE STANDARD RAYTRACING COMPETITION COPYRIGHT.
JPGFILE: warm_up.jpg
ZIPFILE: warm_up.zip
RENDERER USED: 
    MegaPov 0.7

TOOLS USED: 
    Spatch, LParser, Xfrog, Plantstudio,
            3DWin, 3D-Exploration, OBJuvPOV, Warp_s/Colefax_s
            Compressed Mesh Macro, Photoshop

RENDER TIME: 

parse 11 min 40 sec / trace 101 h 18 min (AA 0.1)/ 1110 MB peak memory

HARDWARE USED: 
    1,4 GHz Athlon C / 1 GB RAM



IMAGE DESCRIPTION: 

Morningtime: a swarm of butterflies is warming up near a small arm 
of a river.
Generally butterflies are only able to fly at a body temperature of
30 _C or above. Once they are flying, they can keep up their body
temperature.
Painted Ladies are the most widely distributed butterflies in the world
and can be found on all continents except Australia and Antarctica. 
Occasionally the butterflies make mass migrations in search of a new 
habitat.
And perhaps sometimes a swarm can be seen at a sunlit meadow
warming up...



DESCRIPTION OF HOW THIS IMAGE WAS CREATED: 

The image is centured around the idea of generating complex, but
random landscapes with povray. Thus I was experimenting with
"trace" and "eval_pigment".
These powerful megapov commands make it possible to place an object 
at another object with a probability given by accompanying
imagemaps.
All objects in this image are placed with this method.

The ground consists of two heightfields - a small one in the foreground
and a nine times as large one in the background. The imagemaps were 
first drawn by hand in photoshop and then disturbed by a clouds filter.
The water is a heightfield based on pov code.

There are 181 trees (eight different ones), 55500 grass blades, 3000 
ferns (each with two different sorts) and many other plants like 200 
rotten leafes near the waterline in the foreground. Then there are 500
butterflies placed on the ferns in the right foreground and 1200
butterflies placed "in the air". The objects were placed with nine
different maps.

Most of the trees are created with Xfrog (http://www.greenworks.de/), 
exported as an obj file and converted with OBJuvPOV .
The grass is not mine. The bushy one is based on Mick Hazelgrove_s
grass macro (povray.binaries.scene-files / 14.nov 99) and the other one is 
basically from Samuel Benge_s country_grass macro
(povray.binaries.scene-files / 25.june 00) with contributions from Anthony 
Bennett  (povray.binaries.scene-files / 1.july 00). I added some more
 random factors. The macros can be found at news.povray.org. 
I took these macros because I haven_t had any experience in defining meshs 
up to now.
The great fern is from LParser (http://www.xs4all.nl/~ljlapre/lparser.htm). 
The direct export to povray leads to very slow renderings. But export in a 
mesh formate, conversion, import with OBJuvPOV and compression with 
Warp_s/Colefax_s Compressed Mesh Macro (http://www.geocities.com/
ccolefax/pcm.html) gives good results in terms of time and memory savings. 
All 2500 ferns consume together only about 7 MB. Analogous 
compression of other meshes failed so far because of texture problems.
Rest of the plants are hand coded or made with Plantstudio 
(http://www.kurtz-fernhout.com/PlantStudio/) or Xfrog.
The rotten leafes were created with Spatch.
The butterflies are an adaption of Micha Riser_s swallow tail 
(http://www.povworld.de/cgi-bin/objects/Nature/Animals/).
The code was changed to allow different fore- and backsides of the
wings. All textures were accordingly changed. 

The main problem was memory consuming. Some objects behaved fine and 
consumed only a small amount of memory for the copies. But some
objects, especially the pov coded ones, behaved as if they would need 
an exact multiple of the single object.
As I added more and more plants, my old computer with 224 MB RAM
become more and more instable when rendering all objects.

There was also a problem with the lights. I couldn_t use radiosity
because of the many objects. One night I started a small test render
with only the trees and the most simple radiosity settings thinkable.
Only thousend points were rendered after 16 hours despite parsing fast!
But with only fill lights the shadowed parts of the image looked flat.
So I defined a global illumination fake with several lights evenly
distributed on the sky. I took the coordinates of a buckminsterfulleren
(looks like a football (soccer in USA)). From these 60 coordinates I
choose the 11 most relevant ones and put dim shadowing lights on them.
Render time is six time as long as with fill ligths only, but the
results are worth the effort.

The next problem was that my computer didn_t start to render the final
image with all objects and the global illumination fake. I deinstalled
most programs, transferred unneeded data to other harddisks and
increased the swap space to 3 GB. But it was all in vain. After 20 h
of severe harddisk trashing the computer hasn_t even begun to trace.
So I gave up and bought a new computer with the fastest X86 processor
available today and with 1 GB RAM.
This was a very good idea. Parsing time reduced to 12 minutes (!) and 
tracing was a magnitude faster.
Unfortunately, the time didn't suffice for a correct test any more. 
Because of this some textures seem unharmonious.

The zip file contains all but the tree files because of their size.
I also scaled the imagemaps down to reduce the zip file size
further.

Finally I want to thank all, who made comments regarding my WIP_s on 
news.povray.org.

