Vray Wish: IR map adaptive caustics parameter

[Thomas An._cg]

This is not a VrayForRhino wish, but .... if Vlado is reading this:

It would be nice if the irradiance map had the ability to automatically take more samples (much much more samples) in areas where caustics (reflective/refractive) are supposed to fall... this way we would not have to bump up the settings so much and the rendering would not have to slow down in areas where it doesn't need to.

Maybe an extra parameter (eg "caustic definition") next to the "color, Normal and Distance" parameters

-

[vlado]

Vray Wish: IR map adaptive caustics parameter

It already does that, to some extent, meaning that it will shoot more rays in areas that have greater variance (which is also where caustics typically are).

The problem is, of course, not limited to caustics, but to any difficult lighting situation in general. Registering such situations is difficult with monte carlo methods, as they rely on random samples - meaning that you may need a whole lot of samples just to determine if there are caustics at all. Putting more samples into them, once you know they are there is quite easy... but finding them initially is not. Bi-directional methods help in some limited cases (e.g. directly visible caustics), but are in the same boat when a more difficult situation arises (e.g. caustics in a mirror).

All in all, I've come to the conclusion that monte carlo methods are not powerful enough to handle difficult situations reliably; even MLT, which is considered the absolute weapon for difficult situations, must first find the difficulty before putting more effort into it and this in itself can take quite a lot of samples (and time).

So, I think a more powerful method than raytracing in needed, at least as a prepass for analysing the scene and guiding the monte-carlo sampling later on. This is a matter of constant research

Best regards,
Vlado

[Thomas An._cg]

Vray Wish: IR map adaptive caustics parameter

It already does that, to some extent, meaning that it will shoot more rays in areas that have greater variance (which is also where caustics typically are).

The problem is, of course, not limited to caustics, but to any difficult lighting situation in general. Registering such situations is difficult with monte carlo methods, as they rely on random samples - meaning that you may need a whole lot of samples just to determine if there are caustics at all. Putting more samples into them, once you know they are there is quite easy... but finding them initially is not. Bi-directional methods help in some limited cases (e.g. directly visible caustics), but are in the same boat when a more difficult situation arises (e.g. caustics in a mirror).

I see. Thank you for this insight Vlado !

All in all, I've come to the conclusion that monte carlo methods are not powerful enough to handle difficult situations reliably; even MLT, which is considered the absolute weapon for difficult situations, must first find the difficulty before putting more effort into it and this in itself can take quite a lot of samples (and time).

So, I think a more powerful method than raytracing in needed, at least as a prepass for analysing the scene and guiding the monte-carlo sampling later on. This is a matter of constant research

Research always sounds exciting ! Maybe it is time develop bigger and better things