Ground Projection: Corona vs. V-Ray

[JonasNöll]

Ok, sent! Thanks!

[^Lele^]

Here.
Match projection with camera XY, use "radius" (which should be "height") to match the camera height.
This works for this type of map.
Other ones, for whatever reason (i think it's tripod height on the horizon, but don't quote me on it) need the camera Z into the map's Z position.

p.s.: if you want it more squished, or elongated, adjust "radius" slightly. distortion is cured along X and Y by fixing the position.

[^Lele^]

With Corona, it's a lot easier to see what the projection is doing, so we clear the air of misunderstandings.
Due to the image count it may be more than one post.

First: a false positive. The image looks ok from this angle, but as we turn around, the distortion is visible. The settings are off, and as such the look is *something* arbitrary.
Second: What i deem the right settings. Now it's a match with V-Ray's. The projection is clearly active, the distortion gone, whichever way we look.
Third, a side view of the current settings. As the dome is visible, you can tell it's centered on the camera we'll render from, and camera height matches it.
Fourth, we can see what zeroing X and Y in the map position does to it. Along with making its Z negative. The sphere center moves.
Next post for more.

[^Lele^]

Fifth, This is what looking behind us on the map does, with the settings i suggest. The projection is continuous, indeed indistinguishable from the standard, non-ground-projected one. Yet we have seen it as a hemisphere, the projection *is* in effect, the distortion is minimal.
Sixth, the original settings, which i called false positive: looking behind us the distortion where the projection happens is enormous. We see it from this side because it's the one farther from the center (this "pulls") while on the other there's only moderate squishing.
Seventh, the effect of changing "height": the map squishes towards the top of the sphere for positive values, and towards the flat bottom with low ones. Anything outside of the exact value will produce squishing (i.e. vertically.) and horizontal lines' warping.
Eight: Example of distortion from pulling up the tripod height.

[^Lele^]

To conclude, all of this can be done with either engine.
Radius for v-ray is misnomered, working like height does for Corona (And OSL).
Radius in Corona changes the Dome size, hence the *lighting*, not the projection.
If the radius is set incorrectly, in Corona, the lighting will be very wrong (the dome won't light the scene), something that doesn't happen with us.

[kosso_olli]

Here.
As explained in the five pages above, time and again.
Match projection with camera XY, use "radius" (which should be "height") to match the camera height.
This works for this type of map.

Lele, for the last time: This result is wrong. It is not a ground projection. By typing in these values, you basically make this a spherical projection. The scale on the ground floor is all messed up! Please compare it to the Corona result on page 1. The car is sitting right between the lines in the parking lot, the background is not distorted. While your solution fixes the distortion (by mistake, because I count this as a bug), it will mess up the scaling for the ground floor, because it is not a ground projection anymore. Look where the teapot is standing in your image and what the size of the parking lines is!
This result is not identical to Corona, so that is a plain lie.

[^Lele^]

This result is not identical to Corona, so that is a plain lie.

Really, now.

[kosso_olli]

Really, now.

It just isn't! I don't know where your teapot in V-Ray is standing in the image, but it's not the ground.




Also, somewhere along your tests, you managed to get the same wrong result from Corona as well. No idea what you want to prove, here.

[^Lele^]

Can you not see in Corona that the projection is *always* on, regardless of settings?
The hemisphere it's projected on is seen from the outside, as its the wont of a discrete one, if viewed from a persp view.
I negate distortion of the projection, not the projection itself.
Distortion i show you with the side views and the hemisphere, in the previous images.
You can always fiddle with the position and height, from a neutral projection, and get amenable translation and squishing out of it.
With both: with equal settings, both produced an *identical* undistorted ground projection.

[kosso_olli]

And another hint why this is wrong: On car shootings, the HDRI probe is taken exactly in the location where the car whas standing before (middle of the car around the A-pillar). This ensures that the reflections on the rendered car match as closely as possible, without any stretching or artifacts.
By changing the X-Y-Z-values for the ground projection, you move the center of origin to the in-scene-camera location, as can be seen here:



Teapot is in coordinate origin (where the car would be), but the projection origin is at the in-scene-camera location. As you can see, the area around the teapot is already showing artifacts, and it gets worse the further the car is away from the camera.

You'd want a result like this, where the projection of the environment on the ground matches the location of the on-set car. Much less artifacts, reflections better matching the shot environment.



If that still doesn't make it clear, I will show you production images, with on-set reference and the rendered result of your and my solution. So much work for such a little thing.

[^Lele^]

Well, the pinching is where the camera that took the HDRI was.
The map should match that, as it has no knowledge of the scene, or of where the subject is, which has nothing to do with the projections.

As stated above, feel free to muck about with positions and such, but in doing so both engines produce artefacts, us more than them because our radius is infinite, while theirs isn't.
Other than that, there is truly no appreciable difference that i see.

Maybe others will, i'll leave it to them, and ungracefully bow out of this.

[kosso_olli]

us more than them because our radius is infinite, while theirs isn't.
Other than that, there is truly no appreciable difference that i see.

And that is why something should be done about this.

[^Lele^]

And that is why something should be done about this.

No.
My opinion is that you should set the projection up properly, not bank on minimising arbitrary distortion via radius.
Just so we're clear.

[kosso_olli]

V-Ray tries to preserve the way the dome light looks above the horizon so that the illumination doesn't change; it is still infinitely far away and doesn't have an actual radius. This is different from Corona where the dome light becomes actually a hemi-spherical light.

Best regards,
Vlado

Are there any plans to improve this behaviour? A simple switch for the user to decide which mode to use would be fine.

[Neilg]

Lele, for someone employed at chaos group you are always acting at odds with what was famously great support. The way you respond to people is gross. Have you seriously never been pulled aside and told to dial it back a bit?