[Paraview] On further investigating the VRPlugin

Emiliano Pastorelli hogancito at gmail.com
Sat Mar 1 02:40:01 EST 2014


Hi All,
I managed to achieve an almost optimal configuration of screens and 
tracked devices on a small 3-screens display (through vrpn only yet, I 
use a razer hydra to track the wands interaction and a wintrackerIII for 
the head position).
Now with all of that properly configured, i had a chance to test more in 
detail the vrplugin.

My feeling is that it's based on a couple of wrong assumptions.
One above all, it's based on the origin of the scene. That means that if 
you're moving you're object a bit around (and i've seen most of the 
users do it a lot), at a certain point the head movement goes wrong.
If I try with a simple cylinder built on the origin, and i bring it on 
the 0,0,0 of my system (defined in the cave file in relation to my 
screens potions), then the head movements are correct.
If my move my head to the left, i see the object correctly "rotating" so 
to show me his right side (considering the cylinder facing me) and it 
slides to its own left a bit as it should (therefore also moving 
properly among the screens as well).
But if i push the cylinder back a bit, cause maybe i wanna have a look 
at it in smaller size, then i move the head to my left, the object also 
moves to the left, in a completely wrong behavior.
Shouldnt therefore the 0 of the tracker always be based on the 0 of the 
system and not on the 0 of the scene?

Also, the translation with the wand works well, it's natural to be used, 
but the rotation axis, move with each consecutive rotation of the scene, 
and therefore, if i first rotate on one axis the scene, the second 
rotation is gonna be based on the already transformed axis system, and 
soon it will become completely unnatural (not a single test user i had 
on the system felt confident with it).

I see that not many people are using paraview in VR environments, but i 
think with the proper adjustments it could really become a popular tool, 
specially in low cost VR systems like mine (thanks to the fact that 
doesn't require any expensive lib to run).

One last thing, i'm running paraview on a single 16-cores machine with 4 
Nvidia4000 (one for each screen, each with 2 DLP projectors and one for 
the control displays), and specially when i deal with volume 
visualization,  it tends to be even slower than when i run it single 
screen in my other simple test machine that only has a 4-core and nvidia 
600. Is there any known reason for that?

Best,
H


More information about the ParaView mailing list