I think many people skipped this amazing info at yesterday announcement, but when Mark Zuckerberg was on the stage, he promised that the new GearVR driven by the Galaxy S7, will do Foviated rendering. This is amazing news, and might allow the S7 3D rendering to rival that of the one we get with top of the line desktop computers (using GTX980ti graphic card).
I didn't think that was what he was saying - I thought it was that, rather than streaming all the 360 video 'sphere' at the same resolution it would only stream the part you were looking at in HD. I can't believe the S7 would be able to do actual eye-tracking required for foveated rendering.
"noam23" wrote: The S7 has an excellent camera, with extremely fast focus and extremely well low light sensitivity, so I think this might allow the foviated rendering he was talking about.
But the camera faces away from the user - it's on the back of the phone. How would it be able to track a user's eyes? You'd need cameras inside the headset.
Well, there is the selfie camera which is also extremely good... but I was wrong it's not foviated rendering rather dynamic rendering, I watched it again now.
"noam23" wrote: Well, there is the selfie camera which is also extremely good... but I was wrong it's not foviated rendering rather dynamic rendering, I watched it again now.
Sounds good though, I eagerly await to see it in action.
I think dynamic rendering would be possible in 2015 flagships too. Reason being, these devices can already render the full 360 videos and dynamic rendering would actually reduce the load on processor by rendering only that part of the video where the user is looking (using the Gear VR gyro or phone's). I mean, that is the whole point of introducing it. So that faster rendering could be possible with least load on processors. So, unless S7 has a dedicated hardware chip to render 360 videos (like most flagships have for video decoding), I'm keeping my fingers crossed... :geek:
What he spoke about was dynamic streaming wherein the section(s) of a 360 video you are currently looking at are streamed at higher quality while the parts not visible are streamed at lower quality. It will not be tracking your eyes, only your head. He said the video quality would be about 4x better (whatever that means) and that the bandwidth required would be about a quarter of what it is with traditional streaming. Of course, it should go without saying that when you turn your head to look at another part of the video, it will actually be much worse quality for a short time until the streaming server can send you higher quality data for the region you look at.
This has everything to do with streaming more efficiently, and nothing to do with rendering more efficiently.
