Could make a mini map for the living room (with the humans marked on it) and/or a perimeter alert system to avoid being the subject of practical jokes.
Cool, I was excited to see the articles about your eye tracker last week. I don't have any equipment that would allow me to cut lenses precisely like that (did you use a waterjet cutter or something?). I really wonder how you guys are dealing with the calibration issues I mentioned in my post! Are you somehow tracking the motion of the Rift relative to the head?
This would be idea. The only issue is that to get the full power of eye tracking integrated into the rendering system you need a camera that does at least 1000fps and even that is too slow for some stuff (researchers who use saccades as a behavioral readout in monkey research often use an implanted eyecoil because cameras are too slow). Those cameras are not cheap but maybe with a little math they can get away with using a slower one that costs less.
Ultimately I expect this to happen officially because it is just too powerful an interface and there is so much useful data to be gained (plus scientists would love oculus forever). Probably won't happen in the near future since as Plamer pointed out in one of his talks it turns out avatars only need to look in a probable direction to seem realistic.
That is a brilliant idea to use a hot mirror and IR camera like that!
This is probably today's most accurate technique to use for VR eye tracking, right? And it looks super cheap for any company or project to produce this assembly in-house.
That is a brilliant idea to use a hot mirror and IR camera like that!
This is probably today's most accurate technique to use for VR eye tracking, right? And it looks super cheap for any company or project to produce this assembly in-house.
Very cool - I bet psychologists would also love this, pupil dilation is a pretty good quantitative for mental effort IIRC, and with this set-up you can really control the circumstances.
How hard/easy was it to get that image processing working in Halide?
It was a bit tricky to work with Halide. The Halide code itself is beautiful, but it depends on LLVM which takes forever to build and link, and I ran into some snags with Windows support. I originally wanted to write the whole algorithm in Halide and run it on the GPU, but it turned out that the Halide language isn't expressive enough to implement a Hough transform. I had to write that bit in C and I'm currently only using Halide to do some more straightforward image processing tasks like edge detection filters, blurring, and resampling. It works really well for that though, and performance-wise it runs rings around the naive nested C for loops that I would otherwise have to write.
Thanks, been interested in the language from a distance for a while now.
Can't claim to know what a Hough transform is, but do you think the lack of expressiveness can be fixed without interfering with the tricks that make it work? It's still a language under development after all.
I think the language could be made expressive enough, and the developers thought so too when I asked them about it. I think they're focusing on other things for now though.
A Hough transform is kind of a weird beast; not really your typical image processing problem. For anything that looks like a series of convolutions or other simple image operations, Halide is going to be the fastest thing around; even faster than a series of calls to the Intel IPP library (because of the ability to tune memory access patterns across multiple stages at once).
You don't need to tell me that companies who've been doing eye tracking for years are ahead of my little homebrew project. I know that, obviously. But if I can prevent even one patent that claims "eye tracking like before, BUT IN A VR HEADSET NOW", I'll consider that a win.
You don't have a chance. Eye-tracking has been an integrated part of VR headsets for years, since the 80's, and is well and truly sealed up with patent protection. That there are no consumer-ready, plug and play products on the market today is actually a reflection of this fact, not a consequence of omission. The people with these patents are using them in industrial/medical/military/marketing applications - but the consumer version of this tech is well and truly on its way.
Once the glasses-computing wheels start to turn for consumers, expect to see integrated eye-tracking become a stock standard.
21 comments
[ 3.1 ms ] story [ 85.2 ms ] thread[1] Would you trust a clear screenshot of your eye which could probably used for identification flying around the internet?
But on a more serious note, isn't shining IR directly into your eyes from such close distances hazardous?
(disclaimer: I work for SMI)
Ultimately I expect this to happen officially because it is just too powerful an interface and there is so much useful data to be gained (plus scientists would love oculus forever). Probably won't happen in the near future since as Plamer pointed out in one of his talks it turns out avatars only need to look in a probable direction to seem realistic.
This is probably today's most accurate technique to use for VR eye tracking, right? And it looks super cheap for any company or project to produce this assembly in-house.
This is probably today's most accurate technique to use for VR eye tracking, right? And it looks super cheap for any company or project to produce this assembly in-house.
How hard/easy was it to get that image processing working in Halide?
Can't claim to know what a Hough transform is, but do you think the lack of expressiveness can be fixed without interfering with the tricks that make it work? It's still a language under development after all.
A Hough transform is kind of a weird beast; not really your typical image processing problem. For anything that looks like a series of convolutions or other simple image operations, Halide is going to be the fastest thing around; even faster than a series of calls to the Intel IPP library (because of the ability to tune memory access patterns across multiple stages at once).
Once the glasses-computing wheels start to turn for consumers, expect to see integrated eye-tracking become a stock standard.