So now take this replace a wall in your home with an led screen, now you can have a window to a better reality, where you're standing in your living room, but the wall reflects a devil on your shoulder, or a dragon running around your couch...think of the bedridden sick kids, how much fun they could have...the elderly could see themselves in heaven as only science could show...you combine this with artificial sunlight, ambient wind and olfactory effects--you could go outside, inside.
With iclone 5 soon released I would be interested in facial recognition. After having been introduced to how this gadget works I realize it should quite easily be possible and very accurate?
@FaceBookAutomated My software only works on Linux and Mac OS X, but the OpenKinect project has basically the same stuff, and works on Windows as well.
@okreylos I looked at OpenKinect but found no software that allowed 3d Video capture. Could you point me in the right direction? I already have the drivers working and can get a split screen of the real video with the kinect video.
@ssinn5 Hmm. The 3D video component is a relatively thin layer on top of my VR toolkit. It should be straightforward to port that part to run on top of the infrastructure provided by OpenKinect et al. I'm surprised that nobody over there has done that yet.
Would it be possible to automatically change the point-of-view based on the location of an object? For example, could you wear sunglasses, have the computer recognize/track the glasses, and then have the screen show you what you should be seeing from that point of view?
@zeegeezee Yes, that's view-dependent rendering. My software already supports it, but I'm not yet using the Kinect itself to track the viewer's head position. That's something I want to look into soon.
@okreylos Very cool. Since view-dependent rendering already exists, this might as well: originally I wanted to know if it would be possible to wear a head set that would show you what you would be seeing in a synthetic landscape as you physically move around in real life and have the kinect detect your actual location. If this were paired with a headset using a transparent lcd it would be possible to simply "add" images and objects to "real life"
Wow that’s awesome. I was wondering is it possibly to map the doom character (or Gordon Freeman’s nerd suit) over your imagine and have it mimic your motions? I remember reading about the Japan’s proposing a global holographic world cup broadcast, hacking that signal and adding your own hologram would be some real laughing man (ghost in the shell type) stuff.
Just think about what this kind of technology could do for the movie industry. With thousands of more dollars put into making one high definition camera specifically for this purpose could revolutionize special effects as we know it.
@Freekchild Those higher-def cameras already exist; remember that the Kinect is not new technology, it's existing technology at a new and exciting price point. I fully expect this to pop up in movie and TV productions soon.
(I think this is how CNN did those "holograms" during the 2008 election coverage.)
So I guess the next step is to actually put yourself into Call To Duty 5 [or Jack Bauer -CTU Strike Force] and others across the world see you as yourself.
Thats great to see, what you managed to get using this toy. It's amazing. GREAT JOB!
And question: Would it be very hard to add some algorithm, that fills shadows- regenerates things, that are temporarily hidden? For example on simple layers, like walls or your shirt.
@Ragond Maybe taking an automated 10 second picture, asking the user to stay away from the scenery? then overlay a 3d model of that scene over the 3d video, though the scenery would be static
@crythe1 You mean because of the Hellknight? I appreciate your concern, but unless proven otherwise, I claim it is a Fair Use exception under 17 U.S.C. (p) 107, articles 1, 3, and 4.
Think John Carmack would be proud of you using the hell knight lol. He's into "copy-left" so it seems cool...besides its like an old game.fair use lol
You need 2-3 more Kinect sensors, set in specific areas at specific angles, such that your software can combine the resultant data set in to something with fewer "holes". That being said, what you're able to accomplish with 1 is pretty amazing.
Man this is some awesome stuff, I was really hoping we'd see a bunch of hackers working hard at cracking kinect. Thanks for all the information and software...keep up the great work, I'm subscribing to see what's next!
@RobotNV How embarassing. I'd say the reason that not *every* data member is documented is because writing the software was such a rush job. I'm also sorry about not documenting the fact that the readControl function in USBDevice sets bit 7 of the request type mask internally. That caused some confusion, but luckily no damage.
That's what all yugioh nerds want. So then they can have a duel with the virtual monsters, and then when they watch the video they can actually see em. LOL
I was wondering what if we put 4 camera's in all corners of the room and then hok them to the PC & use the software . So then if you drag you'll be able to see ur self from every angle ?
Keep up the awesome work man. "Ah, I just had some code lying around to draw and animate doom 3 models, so I thought I'd just throw it into my hacked Kinect 3D video software for a quick demonstration." Skillz.
Also, youtube supports 3D now, right? Is it possible to upload 3D versions of these videos? Would it even look right? Perhaps only from the "unrotated" camera view?
Please work for a company with near unlimited funding to perfect this. I want real-time 3D interaction to happen with every video and hopefully movies.
I don't think this is very compute intensive, all you need a is a GPU that supports vertex texture fetch, which is pretty much every one made in the last few years. Then just have a vertex shader that fetches the Z coordinate from the depth map, its pretty trivial really.
Very nice, /excited!! I have a mediocre PC, and it is running Windows 7. So it looks like I will never get to use this software, but I would love to watch some 3d youtube videos.
Here's a thought, is it possible to scan your front then save the presets then scan your back and put both of them together to create a 3d image of yourself?
You could alternatively use liquid crystal "active" shutters in front of each kinect projector to ensure that only one grid is projected at any one time.
The device would then require no hardware modification. Tell your software to ignore alternate frames whilst keeping in sync with the shutters. Video can be streamed at normal framerates, with IR data coming in at half the framerate.
For scans of real-world objects with multiple Kinects, overlapping of the projected IR grid will occur. Modified hardware with IR of different frequencies could theoretically solve this problem as each sensor would only look for its own dots.
Use a candle for quick calibration and synchronisation between the two units
These are great videos. You might want to check out some of the things that Dr Richard Marks did with an early version of the Kinect sensor v=jYHr0I-iFHE at around 5:00
An easy way to remove the shadow behind you is to take a control image first, render the background from that, and then composite the 3D render of yourself in over that image.
For 360 deg room renders, make sure that the second Kinect is not blinded by the first, and that the projections do not overlap
I imagine a future where a console has you 3D map the inside of your house, then you wear the console in a back pack connected to a pair of glasses that act like a display, people all around the world wandering around their homes shooting imaginary monsters.
you should throw in some collision on the kinect recording, and physics + collision on the model. That would be pretty cool, but regardless, this stuff is pretty cool that you have shown us.
Excellent work! In the vein of filling out the entire room I suppose would be to fill in persistent models of dynamic objects as well, such that it could detect when bits were missing (arm half dissapearing) and replace that data with previously recoreded data. It might be nice to see some sort of additional edge detection and exponential linear functions used to smoothe the edges of the black areas to provide a more crisp image. Really great work!
So is it possible to achieve that effect using 2 good quality USB webcams? Since Kinect is just two video cams and amazing software. Or am I missing something?
@RAGpikczers you're missing something. there's also a projector involved, and some non trivial algorithms involved to reconstruct the image. Depth reconstruction from 2 cameras is very complicaded and flawed.
@endianAphones Yeah, depth reconstruction from 2 cameras isn't even stable in the current beta version of human beings we're currently seeing ~6 billion instances of on this planet. Especially buggy on friday and saturday night also. How would we go about running that on a silicon based platform?
@RAGpikczers Your missing tons. Kinect is not 2 web cams. Its one Camera, one IR array emitter, and an IR depth sensor. Microsoft wouldnt make u pay so much for a camera, they arnt sony bro.
@DrakenZA Wow obviously your perceptions toward Microsoft are hugely deviated from the truth. It costs $56 dollars to make the kinect, they are charging $150 for it. So yes they would make you "pay so much for a camera".
Amazing! If you use multiple cameras you have a whole 3d environment without any "shadow". I got so stoked on working with this now, haha. I will subscribe, and see how this goes. Good luck mate!
By streaming the data live of this virtual environment, its possible to support headtracking and stereoscopic 3D. I wait for the first videochat with headtracking and stereoscopic 3D feature. It's like the first phone-call. It's pioneer work.
I honestly expected the first 3d model to be used exactly like this, would be Hatsune Miku, not a doom 3 creature...
Guess I was wrong, I'm not disappointed or anything, I think this is great.
It does beg the question, when will we see the videos of people dancing in 3d with their own Miku models? not long I hope, and possibly even next week.
@MrVardelos Someone just told me that she's an Idoru, in the sense of the William Gibson book of the same title. Now I get it. I think I'm staying away from that. ;)
@okreylos sorry, I didn't reply sooner, I don't have an email notification apparently... I'll just rectify that.
I'm perfectly fine with the doom3 model, that is not the point, I just wanted to tell you that I honestly believed Miku would be the first ever model for this kinda thing.
I am a teacher and can immediately see immense applications of this in the classroom. I have however NO technical knowledge of any of the terminology you have in your descriptive text box above. Can you advise me as to how this could be used by a novice like me? This is incredible and thank you for sharing it.
@afkatws That's hard to answer. From one point of view, the 3D video engine is now a "black box" that's working pretty stably, so you could just get a computer, install the software, hook up a Kinect, and you'd be good to capture people and show them in 3D. Right now you'd need a pretty beefy computer, though.
On the other hand, that by itself does not an educational tool make. How to do that, and what it would take, I can't say. Depends on the precise application, I guess.
If you get the multiple kinects working, you will revolutionize possibly everything, I would purchase a kinect just for this and donate to you just because its surreal, to think we could just no clip through our houses and create some amazing videos. Thanks for your amazing contribution to the scene.
Could you not sync up multiple Kinects and have them flash their lights in sequence and only take the reading from 1 at a time ( the one with the lights on ). If this was done fast enough you could build up a true 3d world.
Could you not sync up multiple Kinects and have them flash their lights in sequence and only take the reading from 1 at a time ( the one with the lights on ). If this was done fast enough you could build up a true 3d world.
Amazing. Now imagine about 2 extra cameras for filling about 90% of black gaps, and for example Move controller for interaction with virtual objects... I wonder how many time would it need?
What I would do next is set up a physics simulation, and handle collisions between the depth data and the 3D models, then start knocking balls around the room.
Hey okreylos, you seem to be a genius and can do anything Jesus can with technology. What do you think Microsoft has planned in the future that we weren't expecting? 3D webchat? What do you think? Btw I want that pet monster, you are the shit sir.
So now take this replace a wall in your home with an led screen, now you can have a window to a better reality, where you're standing in your living room, but the wall reflects a devil on your shoulder, or a dragon running around your couch...think of the bedridden sick kids, how much fun they could have...the elderly could see themselves in heaven as only science could show...you combine this with artificial sunlight, ambient wind and olfactory effects--you could go outside, inside.
reinventedloserben 6 months ago
With iclone 5 soon released I would be interested in facial recognition. After having been introduced to how this gadget works I realize it should quite easily be possible and very accurate?
wwwBorsingPhotoCom 6 months ago
Dude, when you improve this, they could use this for movies and stuff.
checkit230 9 months ago
So none of this will work on windows?
FaceBookAutomated 1 year ago
@FaceBookAutomated My software only works on Linux and Mac OS X, but the OpenKinect project has basically the same stuff, and works on Windows as well.
okreylos 1 year ago
@okreylos I looked at OpenKinect but found no software that allowed 3d Video capture. Could you point me in the right direction? I already have the drivers working and can get a split screen of the real video with the kinect video.
ssinn5 1 year ago
@ssinn5 Hmm. The 3D video component is a relatively thin layer on top of my VR toolkit. It should be straightforward to port that part to run on top of the infrastructure provided by OpenKinect et al. I'm surprised that nobody over there has done that yet.
okreylos 1 year ago
Would it be possible to automatically change the point-of-view based on the location of an object? For example, could you wear sunglasses, have the computer recognize/track the glasses, and then have the screen show you what you should be seeing from that point of view?
zeegeezee 1 year ago
@zeegeezee Yes, that's view-dependent rendering. My software already supports it, but I'm not yet using the Kinect itself to track the viewer's head position. That's something I want to look into soon.
okreylos 1 year ago
@okreylos Very cool. Since view-dependent rendering already exists, this might as well: originally I wanted to know if it would be possible to wear a head set that would show you what you would be seeing in a synthetic landscape as you physically move around in real life and have the kinect detect your actual location. If this were paired with a headset using a transparent lcd it would be possible to simply "add" images and objects to "real life"
zeegeezee 1 year ago
that is so awesome! Doom 3 ftw!
demollyon 1 year ago
you should create a virtual you with a greenscreen! please!
HELLPINAVENGER 1 year ago
@HELLPINAVENGER See most recent video. :)
okreylos 1 year ago
Wow that’s awesome. I was wondering is it possibly to map the doom character (or Gordon Freeman’s nerd suit) over your imagine and have it mimic your motions? I remember reading about the Japan’s proposing a global holographic world cup broadcast, hacking that signal and adding your own hologram would be some real laughing man (ghost in the shell type) stuff.
CONsmith 1 year ago
Comment removed
CONsmith 1 year ago
Comment removed
CONsmith 1 year ago
Have you thought about head mounting the kinect in combination with a pair of hmds?
linkmaster112233 1 year ago
that character is from doom 3
targ27 1 year ago
Thanks dude, so promising!!
arashasadi72 1 year ago
Just think about what this kind of technology could do for the movie industry. With thousands of more dollars put into making one high definition camera specifically for this purpose could revolutionize special effects as we know it.
Freekchild 1 year ago
@Freekchild Those higher-def cameras already exist; remember that the Kinect is not new technology, it's existing technology at a new and exciting price point. I fully expect this to pop up in movie and TV productions soon.
(I think this is how CNN did those "holograms" during the 2008 election coverage.)
okreylos 1 year ago
Finally a real glimpse at the future. Great Work OK. Have you been contacted by MS yet?
37coolidge 1 year ago
incredible, amazing thought.
Have you tried using two or more Kinects to synchronize the 3D environment?
it would be interesting for audio visual production, especially in the use of chroma key would be useful.
Hugs ... If you have more information, contact me ... I became interested in the project!
amadeudesign 1 year ago
@amadeudesign I was having the same thought when I saw what just one Kinect could detect.
plus1exp 1 year ago
Wow. Really enjoying the progression in your work! Looking forward to more videos.
footfunk510 1 year ago
So I guess the next step is to actually put yourself into Call To Duty 5 [or Jack Bauer -CTU Strike Force] and others across the world see you as yourself.
TheLaptopDoctor 1 year ago
So the model and animation are rendered in real time? Is it possible to eventually incorporate shading or lighting based on what the Kinect sees?
Too bad it doesn't have a higher resolution camera.
timerider4 1 year ago
this stuff is fucking awesome.
tehkill3r 1 year ago
can I just plug this kinect cam into my pc, compile and run the software and start using this when Vrui VR toolkit 2.0 is released?
Monitorimatti 1 year ago
@Monitorimatti Yes, as long as you run either Linux or Mac OS X.
okreylos 1 year ago
Thats great to see, what you managed to get using this toy. It's amazing. GREAT JOB!
And question: Would it be very hard to add some algorithm, that fills shadows- regenerates things, that are temporarily hidden? For example on simple layers, like walls or your shirt.
Ragond 1 year ago
@Ragond Yes, it's a bit hard, but it's a logical next step.
okreylos 1 year ago
@Ragond Maybe taking an automated 10 second picture, asking the user to stay away from the scenery? then overlay a 3d model of that scene over the 3d video, though the scenery would be static
OverlordCelarent 1 year ago
This comment has received too many negative votes show
remove or u ll b sued by activision. pls. ur clever but also idiotic.
crythe1 1 year ago
@crythe1 You mean because of the Hellknight? I appreciate your concern, but unless proven otherwise, I claim it is a Fair Use exception under 17 U.S.C. (p) 107, articles 1, 3, and 4.
okreylos 1 year ago 13
@okreylos
Think John Carmack would be proud of you using the hell knight lol. He's into "copy-left" so it seems cool...besides its like an old game.fair use lol
GoldNCase 1 year ago
from here it would be really simple to do some awesome interaction - like picking up the character from Doom 3 and so on.
The resolution of the depth field looks like its actually quite good.
Awesome project!
bArT8472 1 year ago
You need 2-3 more Kinect sensors, set in specific areas at specific angles, such that your software can combine the resultant data set in to something with fewer "holes". That being said, what you're able to accomplish with 1 is pretty amazing.
Mnementh2230 1 year ago
it should be a guy with a gun shooting you , not some inoffensive monster. but its okay, you're at the start
great stuff dude!
can you actually try what i just said?
Sven2Perroy 1 year ago
@Sven2Perroy INOFFENSIVE? It's a freakin' HELL KNIGHT! Show some respect!
Just kidding. Stay tuned.
okreylos 1 year ago 11
@okreylos - LOL. That would be awesome if it hurled a big ball of Goo at you like in Doom. :)
roschler 1 year ago
Man this is some awesome stuff, I was really hoping we'd see a bunch of hackers working hard at cracking kinect. Thanks for all the information and software...keep up the great work, I'm subscribing to see what's next!
Figilswif 1 year ago
Show me Left 4 Dead 2 or GTA 4 using Kinect on the PC and I'm sold.
havoKXT 1 year ago
aw i want a desktop baron of hell :(
gtincan 1 year ago
Holy. Shit. It's f***ing 3d.
kg123able 1 year ago
This has been flagged as spam show
Wow! This must be one of the best commented software I've seen.
Almost every data member is commented. Good job!
RobotNV 1 year ago
Wow! This must be one of the best commented software I've seen.
Almost every data member is commented. Good job!
RobotNV 1 year ago
@RobotNV How embarassing. I'd say the reason that not *every* data member is documented is because writing the software was such a rush job. I'm also sorry about not documenting the fact that the readControl function in USBDevice sets bit 7 of the request type mask internally. That caused some confusion, but luckily no damage.
okreylos 1 year ago
Can't wait until there are ways to sync multiple infrared pulses to obtain data from different perspectives.
Yessj 1 year ago
Pretty soon we'll be playing PC games with a Kinect...
kordman916 1 year ago
Could you in future hook up multiple Kinect cameras, calibrate them and have an image of yourself without the blackness?
philiprawks 1 year ago
youtube.com/user/ag2web
AG2web 1 year ago
this is really awesome, greetings from the lake of constance, germany
rand2 1 year ago
star wars-like?
stevenalvaradoCR 1 year ago
thats very impressive. shouldnt it be possible for you to use another kinect to add a 2nd point of view covering all the shadows (or most of them)?
daftpunk909 1 year ago
That's what all yugioh nerds want. So then they can have a duel with the virtual monsters, and then when they watch the video they can actually see em. LOL
CrazyGamer011 1 year ago
You sir are awesome and talented.
dancingpill 1 year ago
great job man! :D
bangonkalist 1 year ago
Hahha i'm an accountant and i don't understand a shit haahhahaha
But.. u ROCK man hahaha xD
Ixidor13 1 year ago
WOW KEEP UP THE GOOD WORK MAN
DontBplayaHaten 1 year ago
I was wondering what if we put 4 camera's in all corners of the room and then hok them to the PC & use the software . So then if you drag you'll be able to see ur self from every angle ?
nasir246 1 year ago
@nasir246 Watch his other videos, he explains why that isn't possible (as far as he knows).
treborrrrr 1 year ago
german english :D
mrtrob 1 year ago
wow, so many possibilities
LenntBear 1 year ago
Who wants a stupid weebo character there anyway?
MrEduardKhil 1 year ago
Comment removed
bbenefiel 1 year ago
Impressive, now i want to give it a try
madcoretom 1 year ago
wonder how things will look like when they make 5 dimentional.
ss77ss 1 year ago
@ss77ss how do you make 5 dimensional?
bangonkalist 1 year ago
@bangonkalist well i was only guessing. maybe aliens can see 5 dementions. but humans only see 3d.
ss77ss 1 year ago
This is Epic! Microsoft are u seeiung this!? This guy should be on ur payroll Right now!
usparks 1 year ago
This stuff is amazing dude, keep up the good work
Deezl123 1 year ago
Keep up the awesome work man. "Ah, I just had some code lying around to draw and animate doom 3 models, so I thought I'd just throw it into my hacked Kinect 3D video software for a quick demonstration." Skillz.
Also, youtube supports 3D now, right? Is it possible to upload 3D versions of these videos? Would it even look right? Perhaps only from the "unrotated" camera view?
arlojeremy 1 year ago
Please work for a company with near unlimited funding to perfect this. I want real-time 3D interaction to happen with every video and hopefully movies.
AntDX316 1 year ago
THE FUTURE OF PORN OH GOD YES
Rayofpain 1 year ago 5
how can you hook it up to a computer??
does it use a normal usb port?
alhamdanfilms 1 year ago
I don't think this is very compute intensive, all you need a is a GPU that supports vertex texture fetch, which is pretty much every one made in the last few years. Then just have a vertex shader that fetches the Z coordinate from the depth map, its pretty trivial really.
digitaldud 1 year ago
@digitaldud That is correct. My code isn't using vertex textures right now, but it would be an easy addition.
okreylos 1 year ago 2
you sir, are a badass (and a half)
aavindraa 1 year ago
Very nice, /excited!! I have a mediocre PC, and it is running Windows 7. So it looks like I will never get to use this software, but I would love to watch some 3d youtube videos.
seifer93 1 year ago
Comment removed
Fantom2067 1 year ago
This has been flagged as spam show
Великолепно!)) интересно как бы это смотрелось если бы вы использовали 4 кинекта =))
Fantom2067 1 year ago
Comment removed
Fantom2067 1 year ago
this is so amazing
AGreenWallSnipes 1 year ago
Another win for german engineering!
Liface 1 year ago
@Liface kreylos is probably greek?
WheresTheAnyKey1 1 year ago
Did you create the software to control the camera or you dowload it from somewhere?
LuisZunigaT 1 year ago
Here's a thought, is it possible to scan your front then save the presets then scan your back and put both of them together to create a 3d image of yourself?
Ambushcrysis 1 year ago
Se fue al carajo este loco! Cogratulations Man!
lucianodato 1 year ago
Comment removed
Digger581 1 year ago
AMAZING!
Novinte 1 year ago
So damn awesome!!!
ruupi01 1 year ago
JAJAJ QUE CABRON!
XSIRIUSXREO 1 year ago
That is VERY cool!
neerdy5 1 year ago
Sehr beeindruckend, ich hoffe es folgen noch weitere Videos!
tonythebeat1 1 year ago
Xbox live indie games, thats all I got to tell you.
shawnzee102 1 year ago
freekin' sweet!
iridium130m 1 year ago
rule 34
b4u68k 1 year ago 25
@b4u68k rule 35
karlahund 1 year ago
@b4u68k Oh god no
josemlopes 1 year ago
@b4u68k holy shit I never thought reading that would read like such a revelation.....I fear the implications =O
conceitarturo 1 year ago
You could alternatively use liquid crystal "active" shutters in front of each kinect projector to ensure that only one grid is projected at any one time.
The device would then require no hardware modification. Tell your software to ignore alternate frames whilst keeping in sync with the shutters. Video can be streamed at normal framerates, with IR data coming in at half the framerate.
magiccjk 1 year ago 4
PS: The virtual camera is over here!
magiccjk 1 year ago
@magiccjk Yeah, I totally didn't notice that. I was looking at the real camera, and forgot what my own project is all about. :)
okreylos 1 year ago 2
For scans of real-world objects with multiple Kinects, overlapping of the projected IR grid will occur. Modified hardware with IR of different frequencies could theoretically solve this problem as each sensor would only look for its own dots.
Use a candle for quick calibration and synchronisation between the two units
magiccjk 1 year ago
These are great videos. You might want to check out some of the things that Dr Richard Marks did with an early version of the Kinect sensor v=jYHr0I-iFHE at around 5:00
An easy way to remove the shadow behind you is to take a control image first, render the background from that, and then composite the 3D render of yourself in over that image.
For 360 deg room renders, make sure that the second Kinect is not blinded by the first, and that the projections do not overlap
magiccjk 1 year ago
Artoo. Let the Wookiee win.
mechan9 1 year ago
that is porn
nullsub 1 year ago
awesome :)
MrVinca 1 year ago
I imagine a future where a console has you 3D map the inside of your house, then you wear the console in a back pack connected to a pair of glasses that act like a display, people all around the world wandering around their homes shooting imaginary monsters.
crackyflipside 1 year ago
зыбааа
11prizrak 1 year ago
man, this is AWESOME!!!!! +100500
TheDrKso 1 year ago
160fps =P
BrokenBjartur 1 year ago
круто
AmurXRUS13 1 year ago
Жёстко!!! о_О
Zerkacher 1 year ago
you should throw in some collision on the kinect recording, and physics + collision on the model. That would be pretty cool, but regardless, this stuff is pretty cool that you have shown us.
DylanTh3Villian 1 year ago
Really looking forward to what you can come up with, really good job! Subscribed aswell.
Kimilojjan 1 year ago
Excellent work! In the vein of filling out the entire room I suppose would be to fill in persistent models of dynamic objects as well, such that it could detect when bits were missing (arm half dissapearing) and replace that data with previously recoreded data. It might be nice to see some sort of additional edge detection and exponential linear functions used to smoothe the edges of the black areas to provide a more crisp image. Really great work!
theguruofthree 1 year ago
Comment removed
fvschie 1 year ago
So is it possible to achieve that effect using 2 good quality USB webcams? Since Kinect is just two video cams and amazing software. Or am I missing something?
RAGpikczers 1 year ago
@RAGpikczers you're missing something. there's also a projector involved, and some non trivial algorithms involved to reconstruct the image. Depth reconstruction from 2 cameras is very complicaded and flawed.
endianAphones 1 year ago
@endianAphones Oh, thanks for answer. I knew it was too good to be true.
RAGpikczers 1 year ago
@endianAphones Yeah, depth reconstruction from 2 cameras isn't even stable in the current beta version of human beings we're currently seeing ~6 billion instances of on this planet. Especially buggy on friday and saturday night also. How would we go about running that on a silicon based platform?
bjeah 1 year ago
@bjeah I mean stereo computer vision not human computer vision.
endianAphones 1 year ago
@RAGpikczers Your missing tons. Kinect is not 2 web cams. Its one Camera, one IR array emitter, and an IR depth sensor. Microsoft wouldnt make u pay so much for a camera, they arnt sony bro.
DrakenZA 1 year ago
@DrakenZA Wow obviously your perceptions toward Microsoft are hugely deviated from the truth. It costs $56 dollars to make the kinect, they are charging $150 for it. So yes they would make you "pay so much for a camera".
OvenBakedChips 1 year ago
Amazing! If you use multiple cameras you have a whole 3d environment without any "shadow". I got so stoked on working with this now, haha. I will subscribe, and see how this goes. Good luck mate!
Gonamez 1 year ago
@Gonamez but you dont, that most likely wont work.
oBLACKIECHANoo 1 year ago
By streaming the data live of this virtual environment, its possible to support headtracking and stereoscopic 3D. I wait for the first videochat with headtracking and stereoscopic 3D feature. It's like the first phone-call. It's pioneer work.
DragonEmotion 1 year ago
I honestly expected the first 3d model to be used exactly like this, would be Hatsune Miku, not a doom 3 creature...
Guess I was wrong, I'm not disappointed or anything, I think this is great.
It does beg the question, when will we see the videos of people dancing in 3d with their own Miku models? not long I hope, and possibly even next week.
lonergothonline 1 year ago 2
@lonergothonline See, that's how not with it I am. I don't even know who or what a "Hatsune Miku" is. :)
okreylos 1 year ago 6
@okreylos No worries. Some of us are quite very happy with a Doom model over Hatsune Miku. :)
MrVardelos 1 year ago
@MrVardelos Someone just told me that she's an Idoru, in the sense of the William Gibson book of the same title. Now I get it. I think I'm staying away from that. ;)
okreylos 1 year ago 2
@okreylos sorry, I didn't reply sooner, I don't have an email notification apparently... I'll just rectify that.
I'm perfectly fine with the doom3 model, that is not the point, I just wanted to tell you that I honestly believed Miku would be the first ever model for this kinda thing.
no other thought occurred to me at all.
I have a link to a great article on miku.
lonergothonline 1 year ago
@okreylos I hope you research and understand, so that future videos will utilize Hatsune Miku.
mcilrain 1 year ago
@mcilrain I'm glad you have faith in me, but don't put your hopes up too high. ;)
okreylos 1 year ago
this is amazing, good luck to you....... from jerusalem, israel.
segan25 1 year ago
Comment removed
afkatws 1 year ago
I am a teacher and can immediately see immense applications of this in the classroom. I have however NO technical knowledge of any of the terminology you have in your descriptive text box above. Can you advise me as to how this could be used by a novice like me? This is incredible and thank you for sharing it.
afkatws 1 year ago
@afkatws That's hard to answer. From one point of view, the 3D video engine is now a "black box" that's working pretty stably, so you could just get a computer, install the software, hook up a Kinect, and you'd be good to capture people and show them in 3D. Right now you'd need a pretty beefy computer, though.
On the other hand, that by itself does not an educational tool make. How to do that, and what it would take, I can't say. Depends on the precise application, I guess.
okreylos 1 year ago 8
hahahaha, the reveal of the model was so awesome.
Hergonan 1 year ago
If you get the multiple kinects working, you will revolutionize possibly everything, I would purchase a kinect just for this and donate to you just because its surreal, to think we could just no clip through our houses and create some amazing videos. Thanks for your amazing contribution to the scene.
FlukieLukie 1 year ago
Could you not sync up multiple Kinects and have them flash their lights in sequence and only take the reading from 1 at a time ( the one with the lights on ). If this was done fast enough you could build up a true 3d world.
AlienHorde 1 year ago
@AlienHorde well I guess you can power the emitters on one at a time then their cameras and yeah I think that could work.
Hergonan 1 year ago
Could you not sync up multiple Kinects and have them flash their lights in sequence and only take the reading from 1 at a time ( the one with the lights on ). If this was done fast enough you could build up a true 3d world.
AlienHorde 1 year ago
Amazing. Now imagine about 2 extra cameras for filling about 90% of black gaps, and for example Move controller for interaction with virtual objects... I wonder how many time would it need?
Hunti 1 year ago
Great, Great, Greeeeeeeeeeeeeeat. :)
ric84italy 1 year ago
What I would do next is set up a physics simulation, and handle collisions between the depth data and the 3D models, then start knocking balls around the room.
digitaldud 1 year ago
OMG!!!! This is awesome!
Wizartar 1 year ago
Awesome :]
Superpotet 1 year ago
Hey okreylos, you seem to be a genius and can do anything Jesus can with technology. What do you think Microsoft has planned in the future that we weren't expecting? 3D webchat? What do you think? Btw I want that pet monster, you are the shit sir.
MutantKush 1 year ago