 |
|
|
|
|
|
| |
| |
|
|
From: Thomas de Groot
Subject: Eavesdropping WIP - radiosity issues
Date: 14 Sep 2006 11:12:49
Message: <45097171@news.povray.org>
|
|
|
| |
| |
|
|
I need your expert comments :-)
This is the first serious render of the scene with radiosity. It is a 2-pass
job (first gathering radiosity data on an image half the size of this one;
took about 5+ hours; second pass was just about 45 minutes). There are still
ugly artifacts visible, especially on the columns. What is the best way to
go from here? I include the radiosity settings used so far. Normal and media
are off; Megapov settings were used
Thanks for the help!!
Thomas
//==== start code ====
radiosity {
//saving radiosity data:
#if (RadSave)
save_file "Oreille2.rad"
pretrace_start 0.01
pretrace_end 0.005
nearest_count 6
always_sample on
#if (MPov)
samples 1
adaptive 2
error_bound {0.1 adaptive 1.5, 50}
#else
error_bound 0.1
#end
//reading radiosity data:
#else
load_file "Oreille2.rad"
pretrace_start 1
pretrace_end 1
nearest_count 6
always_sample off
#if (MPov)
samples 1
adaptive 1
error_bound {0.1 adaptive 1.8, 20}
#else
error_bound 0.1
#end
#end
//common settings:
low_error_factor 0.5
gray_threshold 0
adc_bailout 0.01/2
minimum_reuse 0.015
count 150
brightness 1.0
recursion_limit 3
normal Nor
media Med
}
//==== end code ====
Post a reply to this message
Attachments:
Download 'Eavesdropping_wip06.jpg' (58 KB)
Preview of image 'Eavesdropping_wip06.jpg'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Is that texturing you added to the columns and tiles or is that the
artifacting? If it is, it almost looks like a nice texture. ;)
Typically I drop the recursion limit down to 1 or 2 and jack the count
setting way up. Although a significant portion of your image is in the
shadow, and with the nooks and crannies in the architecture you are
using, keeping it at 3 is probably best. You might want to try 2 and
see if you can get away with it. The reason you'd want to lower it is
to save time, because your count is way too low. I'd bump up to 1000 at
least, and at a recursion limit of 3 thats going to take forever.
The biggest factor here might be your error bound. A value of 0.1 is
very accurate and requires a very high count (and ages to render).
Maxing out at 1600 you are still going to see artifacts probably. I'd
raise that up to at least 0.5 and you might want to go all the way to 1.
Your lighting won't be as accurate but it will be much smoother.
Remember, as your error bound lowers, the count has to go up.
Nearest count could also be increased to take a few artifacts out.
Skip
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thomas de Groot wrote:
> I need your expert comments :-)
>
> This is the first serious render of the scene with radiosity. It is a 2-pass
> job (first gathering radiosity data on an image half the size of this one;
> took about 5+ hours; second pass was just about 45 minutes). There are still
> ugly artifacts visible, especially on the columns. What is the best way to
> go from here? I include the radiosity settings used so far. Normal and media
> are off; Megapov settings were used
Well, what I usually do when using the two-pass method, is to use a
lower error_bound on the first pass, and double or quadruple (always 2
with an exponent attached) the error_bound on the loading pass. What
this does is average more of the samples together, in effect smoothening
the radiosity results somewhat. Of course this leads to loss of details,
but detailed errors are smoothened away as well. You could have a look
at my website as well, I've experimented a little with radiosity some
time back:
http://www.nolights.de/projects/radiosity/radiosity.html
Hope that helps, even if just a little.
Regards,
Tim
--
aka "Tim Nikias"
Homepage: <http://www.nolights.de>
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Skip Talbot" <Ski### [at] aol com> wrote in message
news:45098a9f$1@news.povray.org...
> Is that texturing you added to the columns and tiles or is that the
> artifacting? If it is, it almost looks like a nice texture. ;)
That's what I wanted to say! Really, I admire Thomas' work, but I too
have no experience with Rad, and consequently, don't know what an 'artifact'
is. When they're spoken of, I always find it hard to see what the author is
seeing.
I actually see this image as an improvement on the last one. It looks
good to me.
~Steve~
> Skip
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thomas de Groot wrote:
> I need your expert comments :-)
>
> This is the first serious render of the scene with radiosity. It is a 2-pass
> job (first gathering radiosity data on an image half the size of this one;
> took about 5+ hours; second pass was just about 45 minutes). There are still
> ugly artifacts visible, especially on the columns. What is the best way to
> go from here? I include the radiosity settings used so far. Normal and media
> are off; Megapov settings were used
I don't understand your use of the MegaPOV adaptive error_bound - if you
use it in the first pass you will loose control over the accuracy of the
radiosity data taken and stored. Furthermore using tighter settings in
the final pass will increase the number of samples taken there (and
using always_sample off at the same time could lead to low quality).
I would try to decrease recursion_limit to 2 and use the saved time to
improve quality if that does not change the appearance too much.
Christoph
--
POV-Ray tutorials, include files, Landscape of the week:
http://www.imagico.de/ (Last updated 20 Aug. 2006)
MegaPOV with mechanics simulation: http://megapov.inetart.net/
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thomas,
I'm curious about this fascinating scene...
lateen rigged caravel
collosal greco/roman pillars
traditional chinese laborer's hat
Where? When?
DLM
"Thomas de Groot" <t.d### [at] internlnet> wrote in message
news:45097171@news.povray.org...
>I need your expert comments :-)
>
> This is the first serious render of the scene with radiosity. It is a
> 2-pass job (first gathering radiosity data on an image half the size of
> this one; took about 5+ hours; second pass was just about 45 minutes).
> There are still ugly artifacts visible, especially on the columns. What is
> the best way to go from here? I include the radiosity settings used so
> far. Normal and media are off; Megapov settings were used
>
> Thanks for the help!!
>
> Thomas
>
> //==== start code ====
> radiosity {
> //saving radiosity data:
> #if (RadSave)
> save_file "Oreille2.rad"
> pretrace_start 0.01
> pretrace_end 0.005
> nearest_count 6
> always_sample on
> #if (MPov)
> samples 1
> adaptive 2
> error_bound {0.1 adaptive 1.5, 50}
> #else
> error_bound 0.1
> #end
>
> //reading radiosity data:
> #else
> load_file "Oreille2.rad"
> pretrace_start 1
> pretrace_end 1
> nearest_count 6
> always_sample off
> #if (MPov)
> samples 1
> adaptive 1
> error_bound {0.1 adaptive 1.8, 20}
> #else
> error_bound 0.1
> #end
> #end
>
> //common settings:
> low_error_factor 0.5
> gray_threshold 0
> adc_bailout 0.01/2
> minimum_reuse 0.015
> count 150
> brightness 1.0
> recursion_limit 3
> normal Nor
> media Med
>
> }
> //==== end code ====
>
>
Post a reply to this message
|
|
| |
| |
|
|
From: Thomas de Groot
Subject: Re: Eavesdropping WIP - radiosity issues
Date: 15 Sep 2006 02:48:55
Message: <450a4cd7@news.povray.org>
|
|
|
| |
| |
|
|
"St." <dot### [at] dotcom> schreef in bericht news:4509911a@news.povray.org...
>
> "Skip Talbot" <Ski### [at] aolcom> wrote in message
> news:45098a9f$1@news.povray.org...
>> Is that texturing you added to the columns and tiles or is that the
>> artifacting? If it is, it almost looks like a nice texture. ;)
>
> That's what I wanted to say! Really, I admire Thomas' work, but I too
> have no experience with Rad, and consequently, don't know what an
> 'artifact' is. When they're spoken of, I always find it hard to see what
> the author is seeing.
>
Yes, a good question Skip and Steve! I am not sure what it is. I use an
averaged texture for the columns and I think that partly it is texture what
you see. But only partly. I suspect them to be mostly artifacts because the
texture without any radiosity or fancy adds looks much smoother, especially
at the middle distance.
> I actually see this image as an improvement on the last one. It looks
> good to me.
>
Thank you Steve! I thought so too. I always try to put radiosity in because
the look becomes more natural. However, the render times and the tuning of
it is really arduous in my experience.
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Christoph Hormann" <chr### [at] gmxde> schreef in bericht
news:eec9lp$llh$1@chho.imagico.de...
>
> I don't understand your use of the MegaPOV adaptive error_bound - if you
> use it in the first pass you will loose control over the accuracy of the
> radiosity data taken and stored. Furthermore using tighter settings in
> the final pass will increase the number of samples taken there (and using
> always_sample off at the same time could lead to low quality).
>
Thanks Christoph! I had not realised the consequences of the adaptive
error_bound in the first pass. Quite simply I thought it would - at the same
time - speed up calculations and increase accuracy! Well well well... Things
are never as simple as one believes...
> I would try to decrease recursion_limit to 2 and use the saved time to
> improve quality if that does not change the appearance too much.
>
Yes, the recursion_limit is an issue I have been thinking about. I had
already tried as a test a value of 1 but that was ugly.
Thanks for your comments. Much appreciated!
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"dlm" <me### [at] addressinvalid> schreef in bericht
news:4509b646$1@news.povray.org...
> Thomas,
> I'm curious about this fascinating scene...
> lateen rigged caravel
> collosal greco/roman pillars
> traditional chinese laborer's hat
> Where? When?
>
He he!! Yes... all that is true. This is an a-historic mixture of elements
from different cultures and different times.
- The ship in the background is, I believe, a fairly recent (perhaps 19th
century) Mediterranean ship;
- The ships at the quay are ancient Greek merchant ships. The model is based
on the one found at Kyrenia (Cyprus).
- Pillars: This is a reconstruction of the Greek temple of Apollo at Dydima
(Turkey). In fact it is in a state it never was in reality, as the
construction was never finished after a last earthquake destroyed part of
the work.
- Chinese hat: yes, or rather: a variation upon the chinese hat;
- one could add: clothings that reminds of the mythical Middle East of
Haroun Al Rachid;
...So.... it is a story, a myth, a dream... whatever you want... :-)
Still, I feel it important that the elements seem to go together, and blend
into a credible scene.
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"Tim Nikias" <JUS### [at] gmxnetWARE> schreef in bericht
news:45098c3c@news.povray.org...
>
> Well, what I usually do when using the two-pass method, is to use a lower
> error_bound on the first pass, and double or quadruple (always 2 with an
> exponent attached) the error_bound on the loading pass. What this does is
> average more of the samples together, in effect smoothening the radiosity
> results somewhat. Of course this leads to loss of details, but detailed
> errors are smoothened away as well. You could have a look at my website as
> well, I've experimented a little with radiosity some time back:
> http://www.nolights.de/projects/radiosity/radiosity.html
>
Thank you Tim, I shall visit again your site. Everything helps. I try to
understand what I am doing especially by studying what others have done,
however with radiosity this is not always easy.
Thomas
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
