Not sure how much stereoscopic effect you can get from the air.
We did the same last year with Oculus Rift from a plane (see below), and it just looked like regular FPV.
You need to be a member of diydrones to add comments!
Not sure how much stereoscopic effect you can get from the air.
We did the same last year with Oculus Rift from a plane (see below), and it just looked like regular FPV.
You need to be a member of diydrones to add comments!
Comments
@Monroe,
Yes, a significant number of people on the dev team are moving towards the ODroid for any image recognition type work. It's close-to-opensource I'd say. The schematics are published but not the gerber or eagle files. The beagle bone black is another good choice.
The ODroid seems to be about 10x ~ 20x as fast as a RaspberryPi but only 1/3rd the power of my desktop (Dell XPS 8500) so it's quick for a small computer but it's not Hercules. Here is a youtube video of an OdroidX displaying the video from 4 web cams but it's not doing any processing of them I think.
Maybe better to just use one camera and a fast gimbal and kill the lag with sufficient processing punch (192 Nvidia Cuda cores for instance.)
Just a thought.
@Tom, yes, love that idea. Rob Lefebvre (tradheli) brought it up recently as well. There are a couple of videos on youtube of people using very very wide angle video cameras (360degrees actually), sending the whole stream down to the ground station and then removing the distortion and allowing multiple people to look in different directions. Also the panono ball is similar although it's a camera not a video camera. I think the issue is just that the quality is not that high.
@Monroe,
That's really looking great. I guess that's a fixed camera for the moment but I guess it doesn't need to be forever. Looking forward to the unveiling.
Rift 2 is supposed to cut down greatly on the lag.
More significant than the stereo effect (which actually isn't much for a human either because our eyes are so close together relative to FPV type distances) is the really wide field of view of the Occulus.
In comparison with large postage stamp view of FatShark Attitude, the wide field of view is totally immersive rather than the looking down a telescope view of other HMDs.
Cleatrly the developer versions of the rift suffer from inadequate resolution to get past the pixel effect at a truly wide field of view.
But they do acknowledge that and should have it much better under control in their eventual (we hope) release version.
I will be ordering a Rift 2 at the end of the month (when check comes) and am sort of looking at the Nvidia Jetson TK1 as an appropriate device for keeping the lag under control.
I think fully immersive FPV is going to be really significant and the Rift is on the fast track to actually achieving it.
This will be true for more than flying things too, virtual reality is about to come back with a vengeance. and augmented reality has some really interesting possibilities too.
Re the ill feeling thing, I think it's from the lag and one way to get over that may be to use attitude info from the camera gimbal (assuming you have an IMU on it) and instead of viewing it full screen in the occulus instead make it appear like you're looking at the inside of a sphere. The image the camera sees would then be projected on the inside of the sphere. The lag that previously made you sick would now only annoy you as you moved your head around the image on the inside of the sphere would slightly lag behind where your head is pointed.
Wasn't very impressed by the Occulus Riftbook version 1 SDK. It was horrible 1980's quality video, with RGB dots the size of grapefruit. Version 2 SDK is still vaporware, but what high market cap vaporware it is.
In the second video it looks like regular FPV, because only one camera was in use, judging from the video. That way you don't get any parallax, which is the point of the oculus rift, it depends on stereopsis to create the effect.
Cool explanation: http://xkcd.com/941/
Depth perception through parallax decreases heavily after 7 meters, after that other cues like occlusion, relative motion of foreground/background become more apparent. Having objects in the foreground before 7 meters does help to reduce ambiguity on the size of the world, thus it helps to gauge real sizes. That's why painters hold their thumb up at an arms length.