SoylentNews is people
SoylentNews
I had an experience with an HTC Vive a couple of years ago, and I'm now considering getting the hardware required to do proper VR.
Obviously, I'd like to play games, but I'm also interested in visualising data (in particular I see that VTK supports OpenVR).
So I was wondering whether anyone in the community here has succeeded in getting this to work under linux, and if they can comment on the hardware required.
I'd be grateful for any insights.
As I understand it, it's best to get 120FPS, otherwise the brain doesn't like it.
I see that system76 has a "thelio major" desktop that can handle a range of NVIDIA cards, but I honestly don't know which would be the minimum that still gets me reasonable performance.
Is it important to have a lot of memory, a lot of cores?
Will I be able to change the level of detail in games to gain in FPS?
Right now it looks to me like I'd need more than 3000 euros for the whole thing (computer+htc vive).
My wife may not approve.
In any case, with the possibility of a second wave of coronavirus in the winter, I'm under the impression a working VR system would be a reasonable addition to the "don't go crazy" activities around the house.
(Score: 2) by hubie on Friday July 17 2020, @11:46AM (3 children)
Writing your own software should be pretty straightforward. All the tools you'd need are in OpenCV and there seems to be a lot of tutorials on how to do them (search "anaglyph opencv stereo" for instance). You probably know this already, but your results will improve, depending upon what kind of lens you're using, if you remove lens distortion. I used to code this kind of stuff up many years ago before most of the OpenCV tools were available, but poking around online it looks like a lot of the hard work is now packaged into very nice libraries. I'm feeling more than a little inspired now to mess around with this again. There is a lot of interesting things you can do by playing with the camera focal length as well as the baseline separation.
(Score: 2) by Rich on Friday July 17 2020, @12:42PM (2 children)
I wasn't aware of any of this, and just spent some time searching for most possible combinations of "opencv" with "stereo", "display", and whatever and didn't find anything. Note that "anaglyph" isn't what I want here, this is about providing one colour picture for each eye from a single monitor. There's boatloads of stuff related to stereo cameras, and computing depth fields from those, but I found nothing about displays. You don't happen to have a better lead?
(Score: 2) by hubie on Friday July 17 2020, @02:14PM (1 child)
No, I'm afraid I'm not of much use on the display side. You might want to check out the site for the stereopi [stereopi.com] project. They've implemented a stereo camera on a little pi and their website has a blog section [stereopi.com] where they go into interesting projects. Perhaps there are display options/ideas to be mined from there.
I did see this arXiv paper [arxiv.org], which might be in the direction of what you're looking for on the code side of things.
(Score: 2) by Rich on Friday July 17 2020, @03:27PM
Thanks for the leads, at least I learned something new
- The arXiv paper uses a display from lookingglassfactory. it's a passive no-glasses 3D display with a 50 degree FOV; the 15 inch model for three grand, the 9 incher start at 600$. Looking at the specs, these probably must be seen to be believed. The "how it works" vaguely mentions 45 separate views in a lightfield, so it might be a bit resolution challenged, or, I fear, worse for my plan, it can't directly work from stereo images (however they seem to have tools to convert stereo images to what they need). Absolutely nothing about active shutters, though.
- The raspi foundation are asshats wrt the cameras, including a dongle chip, but because they couldn't get away with that with industrial customers, their locked-up driver doesn't check the dongle on compute modules, and there's a nice stereo camera setup with a fitting CM3 carrier from a shop called "waveshare".