Rendering Panoramas

Hi List!

I am playing around with the vt-settings of rpict. I have the panotools
installed, and as I had some cylindrical projection renderings (rpict -vtc ), I
was quite happy with the results when viewing them in ptviewer. However, I
don't know which projection type is needed to get the sphere-projection that
ptviewer seams to support as well. Using -vta gives strange results, and the
fisheye limits me to 180 degrees, so I won't be able to turn around?

I wonder if there are others using ptviewer with radiance pictures, maybe
the answer is too simple to find :wink: If not, I will have to investigate this
deeper - I am quite new to panoramas and I am using them only as a tool to get
an overview of the scenes I am currently building.

CU, Lars.

BTW, do you know about good ressources on using make to do rendering? I know
the example files from radsite, and I know the gnu-make documentation, but
is there anybody who has done some more work on this? I am developing my own
makefile templates at the moment, but I have still lots of problems as some
aspects of radiance differ from the standard build process.

···

--
+++ GMX - Mail, Messaging & more http://www.gmx.net +++
NEU: Mit GMX ins Internet. Rund um die Uhr f�r 1 ct/ Min. surfen!

Hmmm. Using make to do your renderings? This is interesting. I
remember Raphael showing some examples of his make-based stuff
at the Workshop, and thought that was interesting. Being a UNIX
newbie I barely understand makefile scripts that *other* people wrote,
never thought of doing my own.

Question: Why aren't you using rad?

rad does a lot of expert-level (program author-level) thinking about
your model for you, and sets a lot of the myriad switches correctly,
based on three global assumptions: "quality, detail, and variability".
(see radsite.lbl.gov/radiance/refer/rad.pdf for the full lowdown.)

I have been building shell scripts around rad and it's working ok for the
moment. I am in the midst of rewriting a script to better serve my
needs, as sometimes the rad rendering pipeline is not what what you
need, but the ability of rad to intelligently set rendering parameters for
you is a powerful one, and can be (ugh, I hate this use of the word...)
"leveraged" in your shell scrips even if you simply use it to determine
those parameters. You could have rad determine the rendering
parameters and pipe it to a "settings variable". Then call rpict or
rtrace or whatever and reference the settings via the @optfile
statement. You can even tweak the settings with the RENDER=
line(s) in your rif file, which can be manipulated by the shell script with
sed, or whatever, and... well you get the idea.

I'm interested to hear everyone's thoughts on the use of make vs.
shell scripts. Since I'm more of a user than a programmer, I
gravitated to shell scripts because they are really just regurgitated
commands, which is the way I tend to think. But if the "next level" is
make, then I'm all ears. What are you all using out there in the real
world?

···

On 27 Nov 2002 at 1:14, Lars Grobe wrote:

BTW, do you know about good ressources on using make to do rendering?
I know the example files from radsite, and I know the gnu-make
documentation, but is there anybody who has done some more work on
this? I am developing my own makefile templates at the moment, but I
have still lots of problems as some aspects of radiance differ from
the standard build process.

=================================
       Rob Guglielmetti
     [email protected]
http://home.earthlink.net/~rpg777

Hi!

I'm interested to hear everyone's thoughts on the use of make vs.
shell scripts. Since I'm more of a user than a programmer, I
gravitated to shell scripts because they are really just regurgitated
commands, which is the way I tend to think. But if the "next level" is
make, then I'm all ears. What are you all using out there in the real
world?

I am using make here for several reasons. First, only using rad would not be
sufficient, as I build documentation (from docbook), convert the model
(dxf2rad) and mappings (ra_tiff, normpat), build octrees that are used as
instances etc with make. And shell scripts tend to get more and more
complicated if you add objects and "features" later, what is what we are
currently doing here. With make, you add simply one more target. And it does
check a lot of requirements for me, e.g. I don't have to run octree if I do a
"make doc".

Besides, I was simply curious in using some of the well known tools in this
context, make is one example :wink:

CU, Lars.

Hi Lars,

Rob has some good tips about using rad to generate rendering options using the OPTFILE setting, and I use this feature a lot, myself. There are many ways to run the rad command, and the man page is the best place to look for understanding these. In its most basic form, it acts like a simple makefile to build an octree, optionally running mkillum, and maintain a set of rendered and filtered pictures.

However, if you're doing a set of runs over different lighting configurations, scene alternatives, etc., you often need a lot more than rad can offer in terms of scene building. I sometimes use a makefile to maintain the various octrees I use based on scene input, then employ a rad input file that specifies only the OCTREE setting -- no explicit scene files. This can also be specified on the rad command line, so a single rad input file can suffice for many scene alternatives.

For generating animations, the ranimate program does a reasonable job, though it really needs updating for running animations over a network now that rsh no longer exists....

With regard to fisheye projections, the -vta option can go up to a 360 degree field of view, where you get a circular image whose circumference is essentially a single value corresponding to the direction opposite the -vd parameter. I don't know anything about ptviewer, or the kinds of projections it uses. You may need to write a script and pass it to rtrace with the -fac (and -x & -y) options to take ray origin and direction on the input and produce a color picture on the output. Using this method, you can get any projection onto a rectangular image that you can describe mathematically.

-Greg

Hi All,

Quick note on ranimate. It should be possible to use ssh instead of rsh by specifying it to ranimate. Ssh does take a little work to set up correctly for trusted logins (ie logins that do not require a password which is what ranimate needs) but it is possible to do and seems to work fine. I have done so in the past for ranimate, but it has been a while. I will forego all the arguments about why you should be using ssh anyway in place of rsh, they are readily available elsewhere...;->

Regards,

-Jack

Greg Ward wrote:

···

Hi Lars,

Rob has some good tips about using rad to generate rendering options using the OPTFILE setting, and I use this feature a lot, myself. There are many ways to run the rad command, and the man page is the best place to look for understanding these. In its most basic form, it acts like a simple makefile to build an octree, optionally running mkillum, and maintain a set of rendered and filtered pictures.

However, if you're doing a set of runs over different lighting configurations, scene alternatives, etc., you often need a lot more than rad can offer in terms of scene building. I sometimes use a makefile to maintain the various octrees I use based on scene input, then employ a rad input file that specifies only the OCTREE setting -- no explicit scene files. This can also be specified on the rad command line, so a single rad input file can suffice for many scene alternatives.

For generating animations, the ranimate program does a reasonable job, though it really needs updating for running animations over a network now that rsh no longer exists....

With regard to fisheye projections, the -vta option can go up to a 360 degree field of view, where you get a circular image whose circumference is essentially a single value corresponding to the direction opposite the -vd parameter. I don't know anything about ptviewer, or the kinds of projections it uses. You may need to write a script and pass it to rtrace with the -fac (and -x & -y) options to take ray origin and direction on the input and produce a color picture on the output. Using this method, you can get any projection onto a rectangular image that you can describe mathematically.

-Greg

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

--
# John E. de Valpine
# president
#
# visarc incorporated
# http://www.visarc.com
#
# channeling technology for superior design and construction