Chambers <ben### [at] pacificwebguycom> wrote:
> Anyway, isn't this essentially radiosity w/o the pretrace, and unlimited
> count?

well, whatever what he is doing, I hope we don't disencourage him.  Keep going
that we eventually may all be thankful for your effort.

The technique shows promise, if not for things other than the much better
handling of anti-aliasing against very bright surfaces (see the windows against
the white ambient sky or the metal surfaces) and also for not having to lose
nights after nights tweaking radiosity settings that never quite get right --
either are too patchy or too smooth to show correct shadows...

So far, I've not been able to do with radiosity what he's done with path
tracing.  Not to mention the stinking AA clipping problems... after several
hours testing and retesting, I'd surely feel glad to drop all the tweaking
altogether.  Leaving the assembly behind in favor of a compiler... ;)

Post a reply to this message

web.47a63501dcec6a9fce60c5620@news.povray.org...
> The technique shows promise, if not for things other than the much better
> handling of anti-aliasing against very bright surfaces (see the windows 
> against
> the white ambient sky or the metal surfaces) and also for not having to 
> lose
> nights after nights tweaking radiosity settings that never quite get 
> right --
> either are too patchy or too smooth to show correct shadows...

That's actually a big selling point of such techniques, particularly in 
architectural visualization where people need large-size, artefact-free 
images with perfect global illumination lighting: it's just cheaper to throw 
some CPUs at a render when needed (or to rent a render farm) than to pay 
someone to spend hours tweaking some settings and figuring out workarounds, 
only to find that one has to brush out the remaining problems anyway. Of 
course, a hobbyist's perspective is a little different, but this is just to 
say that while rendering speed is always welcome, it can be even better to 
be able to set up GI scenes without too much hassle.
In any case, what Fidos did is extremely interesting.

G.

-- 
*****************************
http://www.oyonale.com
*****************************
- Graphic experiments
- POV-Ray, Cinema 4D and Poser computer images
- Posters

Post a reply to this message

Chambers <ben### [at] pacificwebguycom> wrote:
> Anyway, isn't this essentially radiosity w/o the pretrace, and unlimited
> count?
I think it is closer but not completely the same. It is not raytracing (with
shadow ray) but a real integration of light transport for each point of the
picture.

Regrads,
Fidos.

Post a reply to this message

Hello,

A new test with the same scene. The reduce the hemisphere power and add a box
with ambient term in from of the main window.

I stopped a little bit earlier the rendering than the previous one and more
noise is visible.

I think that the interesting thing is the precise soft shadow (the table leggs
for instance).

Any comment wellcome.

Regards,
Fidos.

Post a reply to this message

Attachments:
Download 'prep4company_low_test_2.01.jpg' (122 KB)

Preview of image 'prep4company_low_test_2.01.jpg'

Hello,

I'm currently playing with a more complex scene (the fabulous kitchen of Jaime
Vives Piqueres http://www.ignorancia.org/en/index.php?page=The_kitchen).

I started with the untextured version (normally used for radiosity pass).

As for the previous scene, I removed all the light sources and add a sky sphere
with an ambient term of 350.

Any comment wellcome.

Regards,
Fidos.

Post a reply to this message

Attachments:
Download '1st-pass.04.jpg' (152 KB)

Preview of image '1st-pass.04.jpg'

o_O

<3

Post a reply to this message

"fidos" <fid### [at] wanadoofr> wrote:
> A new test with the same scene. The reduce the hemisphere power and add a box
> with ambient term in from of the main window.

I'll try the box.  The sky_sphere is no good so far.

> I think that the interesting thing is the precise soft shadow (the table leggs
> for instance).

absolutely, yes!  very difficult to reproduce with radiosity.  Perhaps it is
able to reproduce it, but then takes the same time as in path tracing, in which
case path tracing gets the high ground for getting better AA...

the shadows caused by the lamps in the wall are notable too.  Here's a rough
radiosity render which almost gets it.  radiosity is:
      pretrace_start .04

      pretrace_end .008

      save_file "prep4company-wip2.rad"

      always_sample off

      error_bound 0.1

      nearest_count 8

      minimum_reuse 0.006

  max_sample 1.4

 adc_bailout .01/20

    recursion_limit 2

    brightness Radiosity_Brightness

    count 200*3


I'll tweak it a bit more... perhaps lowering the pretrace_end and error_bound
and lowering count! :P

Post a reply to this message

Attachments:
Download 'prep4company-wip2-save_pass.png' (164 KB)

Preview of image 'prep4company-wip2-save_pass.png'

"nemesis" <nam### [at] gmailcom> wrote:
> I'll try the box.  The sky_sphere is no good so far.

It is interesting to see how far you can go.

I have the feeling that radiosity is good for diffuse inter reflection between
objects (and smooth change of brightness). But it seems difficult to catch
correctly the direct light influence (the use of area light in addition to a
sky sphere may solve the problem of the smooth shadows).

Regards,
Fidos.

Post a reply to this message

fidos escreveu:
> It is interesting to see how far you can go.

yes, well, here's my latest attempt.  Rendertime: 5 hours and something 
for 1st pass, some minutes for 2nd pass with higher error bound.  here's 
the settings:

   radiosity{
     #if (Save)
       pretrace_start .04
       pretrace_end .0008
       save_file "prep4company-wip3.rad"
       always_sample off
       error_bound 0.02
       nearest_count 12
       minimum_reuse 0.008*2
	adc_bailout .01/20
     recursion_limit 2
     brightness Radiosity_Brightness
     count 40*3*5
     #else
        pretrace_start 1
       pretrace_end 1
       load_file "prep4company-wip3.rad"
       always_sample off
       error_bound 0.14
       nearest_count 1
       minimum_reuse 0
	max_sample .4
	adc_bailout .0006//20
     recursion_limit 1
	count 1
     brightness Radiosity_Brightness
     #end
   }


I think I'll next bump up the error_bound slightly to get better blends 
in the first pass.  I'm still with problems getting that second pass 
alright:  there are black-dot artifacts in the corners when I use 
error_bound values close to the 1st pass.  Perhaps anyone more 
knowledgeable has any clue about the settings?

At least, it's smooth, features visible shadows for the lamp in the 
wall, for the top left window corner, for the power outlet in the wall 
and even some shadow near the legs of the table.  Unfortunately, 
rendertime is too high -- perhaps even equalling to path tracing or more 
-- and there are shadow artifacts in the front wall as well as the 
horrid AA light levels clipping issues.

> I have the feeling that radiosity is good for diffuse inter reflection between
> objects (and smooth change of brightness). But it seems difficult to catch
> correctly the direct light influence

you're very right.

> (the use of area light in addition to a
> sky sphere may solve the problem of the smooth shadows).

I still may try that as a last resource. ;)

Post a reply to this message

Attachments:
Download 'prep4company-wip3-save_pass.png' (68 KB) Download 'prep4company-wip3.png' (151 KB)

Preview of image 'prep4company-wip3-save_pass.png'

Preview of image 'prep4company-wip3.png'

oh, and BTW, it seems I forgot to save the latest 1st pass image.  That 
one was from my previous attempt, at a lower resolution.  It looks about 
the same, only smaller and less time consuming.

People were saying that radiosity calculations are independent of screen 
resolution, but that's not what I've experienced:  when rendering the 
second pass at a higher resolution, results sucked very much...

Post a reply to this message