[osg-users] Oculus+OSG
Björn Blissing
bjorn.blissing at vti.se
Tue Sep 1 07:31:26 PDT 2015
Hi Christian,
Well, I don't know how the OSG-DirectShow plugin works regarding threading and latency.
When I did a similar setup I used OpenCV and wrote a plugin that spawn a thread for image acquisition. Here is my code:
https://github.com/bjornblissing/osgopencv
Hopefully the new Oculus SDK has cut the latency a bit, BUT this does not matter much if the main part of the latency comes originates from the image
acquisition.
Best regards
Björn
________________________________
Från: osg-users <osg-users-bounces at lists.openscenegraph.org> för Christian Buchner <christian.buchner at gmail.com>
Skickat: den 1 september 2015 14:57
Till: OpenSceneGraph Users
Ämne: Re: [osg-users] Oculus+OSG
In this case I am using a slightly modified DirectShow OSG plug-in, running the osgviewer in single threaded mode. The plug-in provides an osg::ImageStream and updates a texture that is displayed in the OSG scene. The camera is an USB 3.0 model, and I would expect it to be one of the fastest "web cams" out there.
I am hoping that OpenGL applications also benefit from the low level driver support created for the Oculus and similar VR devices.
Christian
2015-09-01 12:48 GMT+02:00 Björn Blissing <bjorn.blissing at vti.se<mailto:bjorn.blissing at vti.se>>:
cbuchner1 wrote:
> Can you make any statements regarding latency improvements upgrading from the 0.6 SDK to 0.7 for osgoculusviewer based applications?
Hi Christian,
I have not had time to do any measurements regarding latency differences between 0.6 and 0.7 yet. Sorry.
cbuchner1 wrote:
> Our current use case is currently, attaching an Intel RealSense camera to the front of the Oculus DK2 and streaming a live image into the viewer's eyes. Our latency is currently in the area of about 3-4 video frames @75 Hz, amounting to 40-50 ms. This is a bit too high and causes some motion sickness and slow reactions in our unfortunate test persons. ;)
Are you talking about pure rendering latency (motion to photon) or the camera to HMD latency (photon to photon)?
If you are talking about the latter 40-50 is quite good I would say. And as Jan says, the camera is probably the limiting factor in that case. Another problem can be if you run the camera capture in the same thread as the rest of the rendering. Moving the camera capture to an own thread gave me a great performance boost, but results in the cameras running asynchronous with the rendering. This in turn can gives a larger variance in latency.
Best regards
Björn
------------------
Read this topic online here:
http://forum.openscenegraph.org/viewtopic.php?p=64975#64975
_______________________________________________
osg-users mailing list
osg-users at lists.openscenegraph.org<mailto:osg-users at lists.openscenegraph.org>
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
More information about the osg-users
mailing list