This would probably solve the neck injuries caused when VR fanatics wear VR headsets for too long and cause injury due to moving the center of mass of the head away from the neck. Definitely looks a lot healthier. Who's to say though, let's get this to a prototype product and start testing in an industry where it can be tested under careful conditions.
Cool stuff, but the FPS leaves a bit to be desired.. 0.0008333 FPS (1 frame per 20 minutes) on a RTX 3090. Also the FOV etc. of course. Edit 2: To be fair, real time calculation is outside the scope of the paper. Edit: from the paper: " The calculation time of a single phase pattern generation with SGD-CITL and HOGD-CITL for 500 iterations was 10 minutes and 20 minutes with the PC, respectively." If I understand correctly, that's 20 minutes per color, so 60 minutes for RGB. I don't doubt some AI magic or something can bring it down in addition to GPUs becoming petaflop devices in not _too_ long. But if not AI, it requires 50% perf increase per year for about 31.6 years to reach 100 FPS for HOGD-CITL. Also, that's with the current display size and limited FOV. I hope I'm wrong. To get 120° FOV with decent quality a 5cm diagonal SLM with a pixel pitch of 1um is required. I'll make it easy for myself and assume a square aspect ratio. That would be a 35.4mm by 35.4mm SLM with a resolution of 35400x35400 pixels. Yikes. 1253.160 megapixels. Edit 3: I'm not disparaging the paper, just hoping to provide some realism for people expecting this to be in the stores in a few years. If I have misunderstood anything, please let me know.
I'd love for this tech to go to market in the form of lighter, wireless headsets. If NVidia could get to work on an open standard for transmitting display signals wirelessly, that'd be great, especially if the transmitter is a product that can just plug in to an existing GPU's displayport.
Very nice, would work indoors or outdoors if Hologram glasses where thin but blocked light , And had cameras showing more views then the eyes could see. Wide view cameras with telescoping camera's, are even on the best smartphones allready. We just need a few designs with camera bumps like forehead camera's connected, something with more views people would always use. Warnings for moving objects and so on as well.
I watched the video twice to understand the presented tech better and it's very interesting. Would this be more closer to a projector or a CRT display? I'm not sure what is coloring the screen, is it the laser or is the laser hitting a LCD like color filter? Would the response rate and density of detail be comparable or better to the displays you are comparing you solution with?
That looks amazing. I couldn't get my head around this though :D, with this many surface bounces before reaching eye, wont the brightness be incredibly low? I sounds like the micro oled display will have to be extremely high nits?
@@cybyrd9615 I'd go 6 years max and this should be feasible. There are many individuals, indie companies, and corporations working on this already. Also we have no idea what the US military has been working on or has completed. We all know that lightweight AR optics have been working on for years already. Also, I'm going to get shit for this, I always do. FB (now Meta) is already working on upgraded lighter VR headsets. I believe, may need to check again, they are working on 4 different models for their next release. I know people hate on Meta, but you got to admit, they have the financial backing to create something remarkable.
This is a fascinating development. But I imagine a head mounted unit would still need to have the display somewhere to feed the waveguides. This would still require a (potentially much more comfortably placed) display unit somewhere on your head... Or is the plan to have the image source entirely outboard, using very long guide cables?
I really enjoyed this video, however I'm completely lost on that pupil replicating wave guide thing, and how the light is supposed to interact between it and the SLM (SLM? a bit vague in the paper), at least clearly enough.
I don't how exactly the SLM works, but I assume it displays the diffraction pattern which could be activated as a 3d image of the target object, by a laser of particular wavelength, is that what it does? If so, how bad is the workload? Computing a light field isn't easy, or is it?
I think so too.. According to the paper a RTX 3090 took 20 minutes for a single phase pattern generation with 500 iterations of HOGD-CITL. I think that's for one of the color components, so 60 minutes for a RGB "image"
@@phizc Sounds not very promising. If anyone can create a proper movie with the tech it would be the next-gen Avatar, in the sense of how evolutionary it would be.
I can see why Nvidia teamed up with stanford on this one. Those calculations are probably very taxing so some dedicated GPU probably made quick work of it lol
According to the paper a RTX 3090 took 20 minutes for a single phase pattern generation with 500 iterations of HOGD-CITL. I think that's for one of the color components, so 60 minutes for a RGB "image"
I Hope it will not end as other similar projects... NVIDIA already had some good ideas and the implementation never came to the market (as for example stacking LCD panels to significantly increase resolution..)
who cares about a holographic image for a single eye? 2d images with proper offsets for two eyes is indeed 3d. what benefit is there for hologram for a single eye? can you actually do accommodation like that? EDIT: oh shit - you CAN do accommodation?! that's pretty cool.
you look like a type of guy that watches nature documentary then says 'booooooring... i wanna see animals with my eyes' like this is literally just a report, if you dont like it, you can just ignore it or say 'this is not that efficient or marketable design' not say choosy things like 'i wanna get it so you should release it under $400 or im not interested'
Thanks for the likes and comments everyone! Please check our recent “Holographic Glasses Q&A” video too! ruclips.net/video/SPyX0NI-l5E/видео.html
Im a smooth brain but I can appreciate the insane levels of intelligence and work needed to do this. Incredible work
great work team!
This would probably solve the neck injuries caused when VR fanatics wear VR headsets for too long and cause injury due to moving the center of mass of the head away from the neck. Definitely looks a lot healthier. Who's to say though, let's get this to a prototype product and start testing in an industry where it can be tested under careful conditions.
Amazing work! Can't wait to upgrade from OG Oculus
Cool stuff, but the FPS leaves a bit to be desired.. 0.0008333 FPS (1 frame per 20 minutes) on a RTX 3090. Also the FOV etc. of course.
Edit 2: To be fair, real time calculation is outside the scope of the paper.
Edit: from the paper: " The calculation time of a single phase pattern generation with SGD-CITL and HOGD-CITL for 500 iterations was 10 minutes and 20 minutes with the PC, respectively."
If I understand correctly, that's 20 minutes per color, so 60 minutes for RGB.
I don't doubt some AI magic or something can bring it down in addition to GPUs becoming petaflop devices in not _too_ long. But if not AI, it requires 50% perf increase per year for about 31.6 years to reach 100 FPS for HOGD-CITL. Also, that's with the current display size and limited FOV. I hope I'm wrong.
To get 120° FOV with decent quality a 5cm diagonal SLM with a pixel pitch of 1um is required. I'll make it easy for myself and assume a square aspect ratio. That would be a 35.4mm by 35.4mm SLM with a resolution of 35400x35400 pixels. Yikes. 1253.160 megapixels.
Edit 3: I'm not disparaging the paper, just hoping to provide some realism for people expecting this to be in the stores in a few years. If I have misunderstood anything, please let me know.
amazing quality as well, this could be revolutionary
great work! so excited for the future
I'd love for this tech to go to market in the form of lighter, wireless headsets.
If NVidia could get to work on an open standard for transmitting display signals wirelessly, that'd be great, especially if the transmitter is a product that can just plug in to an existing GPU's displayport.
Build it in I think.
Very nice work!
Thanks :)
Pretty cool that they got Mr Krabs to demo it
Top comment
Very nice, would work indoors or outdoors if Hologram glasses where thin but blocked light , And had cameras showing more views then the eyes could see. Wide view cameras with telescoping camera's, are even on the best smartphones allready. We just need a few designs with camera bumps like forehead camera's connected, something with more views people would always use. Warnings for moving objects and so on as well.
디자인이 재밌네요ㅋㅋ
Very impressive
In a few years it might be a vr contact lense. Looks great.
Great work I worked on the first Sony sdg and this way much better design principal
I watched the video twice to understand the presented tech better and it's very interesting.
Would this be more closer to a projector or a CRT display? I'm not sure what is coloring the screen, is it the laser or is the laser hitting a LCD like color filter?
Would the response rate and density of detail be comparable or better to the displays you are comparing you solution with?
That looks amazing. I couldn't get my head around this though :D, with this many surface bounces before reaching eye, wont the brightness be incredibly low? I sounds like the micro oled display will have to be extremely high nits?
Yea itll take a decade for this it work
@@cybyrd9615 I'd go 6 years max and this should be feasible. There are many individuals, indie companies, and corporations working on this already. Also we have no idea what the US military has been working on or has completed. We all know that lightweight AR optics have been working on for years already. Also, I'm going to get shit for this, I always do. FB (now Meta) is already working on upgraded lighter VR headsets. I believe, may need to check again, they are working on 4 different models for their next release. I know people hate on Meta, but you got to admit, they have the financial backing to create something remarkable.
This bouncing is also how fiber optic cables work. The losses can be pretty minimal, hopefully.
@@cybernerdpc2554 oh they better, I have an idea where the 2 key words can't be found together in one patent, the military better have classified it
Get funding. I am staying away from VR until they are light weight, coherent, and holographic. You’ve got all three.
This is a fascinating development. But I imagine a head mounted unit would still need to have the display somewhere to feed the waveguides. This would still require a (potentially much more comfortably placed) display unit somewhere on your head... Or is the plan to have the image source entirely outboard, using very long guide cables?
Walkman sized LED boxes to wear as a waist gadget maybe?
I really enjoyed this video, however I'm completely lost on that pupil replicating wave guide thing, and how the light is supposed to interact between it and the SLM (SLM? a bit vague in the paper), at least clearly enough.
bu şey mi gözlüksüz 3d ekranların yani iç içe geçmiş ekranların lens haline gelip VR gözlük üzerinde daha az yer kaplayan versiyonu mu ?
I don't how exactly the SLM works, but I assume it displays the diffraction pattern which could be activated as a 3d image of the target object, by a laser of particular wavelength, is that what it does? If so, how bad is the workload? Computing a light field isn't easy, or is it?
I think so too..
According to the paper a RTX 3090 took 20 minutes for a single phase pattern generation with 500 iterations of HOGD-CITL. I think that's for one of the color components, so 60 minutes for a RGB "image"
@@phizc Sounds not very promising. If anyone can create a proper movie with the tech it would be the next-gen Avatar, in the sense of how evolutionary it would be.
Interestingly Noice 👍 The fov seems very small though
Yeah, not gonna hold my breath for these to be ready in my lifetime :/
This is what we need! How can we support this project?
Put a highpass filter on your mic audio before rendering.
I can see why Nvidia teamed up with stanford on this one. Those calculations are probably very taxing so some dedicated GPU probably made quick work of it lol
According to the paper a RTX 3090 took 20 minutes for a single phase pattern generation with 500 iterations of HOGD-CITL. I think that's for one of the color components, so 60 minutes for a RGB "image"
It’s BACON STRIPS!
I Hope it will not end as other similar projects... NVIDIA already had some good ideas and the implementation never came to the market (as for example stacking LCD panels to significantly increase resolution..)
I remember this, wish they tried to produce something of that tech, even as a proof of concept piece.
I've solved the problem of the silly looking ribbons that look like rabbit ears.
A hat
That shit is so complicated, to the point where Fourier’s formula took part? Jesus
They fr called the dude from deus ex just for this demonstration
Cool, keep advancing technology.
Is this the same way TiltFive and Magic Leap work?
Thank you.
God bless, Rev. 21:4
完全にSLIのブリッジケーブルぶら下げてて草
After watching this now i know how stupid am i
next time use popfilter please
who cares about a holographic image for a single eye? 2d images with proper offsets for two eyes is indeed 3d. what benefit is there for hologram for a single eye? can you actually do accommodation like that?
EDIT: oh shit - you CAN do accommodation?! that's pretty cool.
I'm a little confused, but they could have made thin vr years ago. Google glasses were thin.
Too bad an SLM costs about $10,000 right now...
What did you just say??
You can keep your virtual world. This will be the demise of the human species.
okey, now work on touch-sensitive gloves
Holoprojectors are here now. This is old hat. No need for glasses
booooriiing .. come back when you have this implemented in a vr headset under 400 usd $ .
smoothbrain goes *ugh ummm laaame*
you look like a type of guy that watches nature documentary then says 'booooooring... i wanna see animals with my eyes'
like this is literally just a report, if you dont like it, you can just ignore it or say 'this is not that efficient or marketable design' not say choosy things like 'i wanna get it so you should release it under $400 or im not interested'