It has been announced that Oculus VR as bought The Eye Tribe, a company known for its eye tracking hardware and software technologies, earlier in December. Facebook’s VR division does not say how it plans to use the acquired company, but the two primary reasons for the takeover could be The Eye Tribe’s gaze tracking IP along with the company’s experience with foveated rendering. Theoretically, the buyout may indicate that foveated rendering could become an important feature of the upcoming VR platforms.

The Eye Tribe was established in 2011 with the aim to develop software and hardware technologies for eye tracking. By 2013, the company introduced its eye-tracking device for PCs and tablets that cost only $99 along with an SDK to develop software for the product. The company actually started shipping the units to developers in 2014, but the devices have never been integrated into actual tablets or computers shipped to end-users. Eventually, the company designed eye tracking technologies for smartphones and earlier this year demonstrated an eye tracking tech for VR headsets.

Eye tracking itself could enable a more intuitive control for virtual reality interactions when combined with gestures, voice and other methods. However, this might not be the only reason why Oculus VR was interested in The Eye Tribe. Usage of gaze tracking also enables developers to use foveated rendering, which allows optimizing the use of GPU horsepower. The method is relatively simple on paper: each scene is rendered first in low resolution and with the least amount of polygons and then the specific part of the screen where the user is currently looking at is rendered in high resolution. Developers may play with such effects like depth of field, blur and antialiasing to further optimize performance and/or improve quality, but all of those tricks require gaze tracking.

The foveated rendering technique is already supported in Unity3D engine and is enabled by the Tobii EyeX tracking device. According to the developers, usage of the tech helped to improve performance of a laptop from 11 FPS to 42 FPS. Microsoft and NVIDIA are also researching foveated rendering and the latter has even come up with new ideas on the matter. The company says that foveated rendering per se is not a panacea for VR (at the end of the day, you can improve performance of any platform by using foveated rendering with gaze tracking) because rendering in lower resolution for peripheral vision results in flickering if the foveation is too aggressive, meanwhile blur itself reduces contrast and introduces a sense of tunnel vision. NVIDIA has a number of suggestions how to minimize negative effects of foveated rendering (e.g., by introducing contrast preservation), but currently this is a research project.

For VR applications, foveated rendering makes sense, especially if we are dealing with a headset based on a smartphone that has a limited amount of GPU horsepower. To enable it, Oculus VR needed gaze tracking IP and a team that understands how things work here. This does not mean that foveated rendering will be a part of Oculus VR’s next-gen VR headset because its current form requires some additional work. However, with eye tracking IP and appropriate team, Oculus VR could implement the technology in the future.

Related Reading:

Source: TechCrunch

Comments Locked

17 Comments

View All Comments

  • edzieba - Saturday, December 31, 2016 - link

    "HMD have been around for a long time, the only difference is now they are small and affordable enough to go mainstream."

    Prior to the current crop of HMDs starting with the DK1, a consumer HMD had:
    - No tracking
    - A tiny (40° for the HMZ series, typical FoV was ~20° for something like the ST1080) field of view
    - Low refresh rate
    - Full-persistance panels
    - Often very low resolution, sometimes even driven entirely by a composite video connection

    And guess what, those cost in the region of $1000 MSRP too! People who have just seen the recent marketing for VR HMDs have no clue the sea-change in the HMD marker Oculus brought,
  • ddriver - Saturday, December 31, 2016 - link

    LOL, you don't need "intellectual property" to put a "larger display" in a HMD, all you need is higher pixel density. There was head tracking in HMDs at least a decade back. Oculus neither invented "higher resolution" nor produce any actual displays. That was enabled entirely by technological progress, and oculus had absolutely nothing to do with that happening, it didn't speed things up, it just popped up when technology advanced enough to make it mainstream.

    As for any "contributions" that it may have brought, those mainly have to do with hype that still fails to materialize into something other than further filling a few guy's pockets. That's what they all do, pick things that are "new" only because they were previously too expensive to go mainstream, and hype the hell out of them, to milk naive and easy to impress dummies, and in the end we aren't any closer to innovation and in a short while it becomes yet another short-lived fad. A few years ago the same thing happened with 3d printers, they were supposedly make a revolution, today nobody cares anymore, they don't even make the headlines.
  • Zan Lynx - Saturday, December 31, 2016 - link

    I used an old Amiga based VR system a few times. It was awful. The Oculus system may not have invented much but the way it is put together works amazingly well. It is not awful.
  • ddriver - Sunday, January 1, 2017 - link

    You don't say, something made today is better than something made decades ago? Who would have thought! Let's hope in a few decades you will also be able to get the point.
  • PseudoKnight - Friday, December 30, 2016 - link

    Don't forget that pupil tracking allows for a more accurate projection, particularly for closer objects. When the pupil moves, it changes the position in which light is entering the eye. I think it was Carmack that mentioned this and how it was part of the ultimate goals of VR HMDs.
  • edzieba - Saturday, December 31, 2016 - link

    I'd go out on a limb and say that we are likely to see lens-compensation appear in a HMD before we see foveated rendering. The performance requirements for foveated rendering in a HMD are harsh and current systems do not meet them, while lens-compensation has a much more graceful failure mode (i.e. the worst failure is the current status-quo).
  • aetrha - Monday, January 2, 2017 - link

    Facebook will of course use eye tracking tech to run "big data analysis" of how people move their eyes, and place even more intrusive and sublime ads everywhere. That's worth billions.

Log in

Don't have an account? Sign up now