Aven in stereoscopic 3D

Pedro Silva Pinto pedrospinto at netcabo.pt
Tue Jul 3 01:29:45 BST 2018

200.0 instead of 20.0 makes it much more acceptable to look at. Still, as I
wrote before, all objects are rendered closer than screen plane.
Maybe some camera positioning could do the trick:
If you at some point try to achieve the rendering of deeper objects farther
away than screen plane and shallower ones closer than it, it will be easy to
see without any glasses: the (left and right) lines will appear to cross as
now they appear almost parallel.


-----Original Message-----
From: Olly Betts [mailto:olly at survex.com] 
Sent: 30 de junho de 2018 23:15
To: Pedro Silva Pinto <pedrospinto at netcabo.pt>
Cc: survex at survex.com
Subject: Re: Aven in stereoscopic 3D

On Sat, Jun 30, 2018 at 12:06:25PM +0000, Pedro Silva Pinto wrote:
> --stereo=buffers is working, but left and right images are way too 
> much separated, causing a lot of eye strain: the 3D image pops out of 
> the screen so much that you'd have to be cross-eyed to see it 
> comfortably! :) Way too close for comfort. Otherwise, the 3D effect is OK.

It seemed OK in the cardboard viewer, but I struggled to make the coloured
glasses work (but they're cereal packet ones so cheap filters with a lot of
colour bleed).

Currently the eye separation is the focal length divided by 20:

    const double APERTURE = 50.0;
    const double FOCAL_LEN = 70.0;
    const double EYE_SEP = FOCAL_LEN / 20.0;

It's possible I cranked that up to make it easier to see it was actually
working correctly and didn't restore the original values.  I'll have to
check where I got that ratio from.

> If the image is moved back closer to screen level, it will be perfect. 
> I think the best practices in 3D rendering advise to set the screen 
> plane halfway along the Z-axis span: closer parts of the cave render 
> in front of the screen plane and more distant parts render 'behind' the
screen plane.

OK, I'll have to think about what that means in terms of the values above.

Meanwhile, if you could try increasing the 20.0 above and suggest a value
which works better than would serve as a useful cross-check.


More information about the Survex mailing list