[Radiance-general] Re: Thesys target

[Axel Jacobs]

Dear Lucio,

> So ... it exists a tachnical report from  CIE which is titled "Test cases to asses the
> accuracy of lighting computer programs". Briefly, it contains some sets of
> scenes well described to be reproduced and computed in any software..

> If nobody did those tests using radiance, it could be a nice goal for my thesys
> to complete them or some of them at least.. If somebody already did them,
> I'd like to look at them, if possible.

Funnily enough, I was actually going to give this very task to one of
my students this year. I think it's well worth doing as an exercise.

There is a validation of Relux (and I think one other software
package) based on the CIE technical report. If I remember correctly,
Relux worked all right, but they discovered that the CIE overcast sky
definition was implemented incorrectly in Relux, which it's now fixed.
If you cant' find  the report on the Net, get back to me, and I'll try
to dig it out.

There are two things that you should be aware of with such a validation:

a) The test cases are all VERY basic. This is because the reference
results are derived analytically (pen on paper). This is to make them
100% correct, but the downside is that the scenarios can't be very
complex. Radiance, however, excels exactly in such cases difficult
cases which can only be validated though elaborate measurements in
real spaces (See J. Mardaljevic, C. Reinhard)

b) Much of Radiance's accuracy is based on the operator's experience
and him setting the ambient parameters correctly. I feel this would
have to be quite a large part of this assessment: How to get them
right, and the penalty of not getting them spot-on.

Those are my thoughts on the subject. I am looking forward to reading
your report, and wish you best of luck with it.

Axel