2 Kinects 1 Box

So just to confirm, in this video he's made a ...3D representation of his 'room' of what the two cameras can see and merged the image together? Hence his ability to rotate he's not actually rotating the cameras but he's rotation the 3D space he's made based of the image being received by the two cameras? The black spaces represents areas the two cameras combined are unsure about either in surface-volume or they have no detail regarding?

Otherwise....he's rotating a camera and switching between camera 1/2?
 
Cool a virtual camera in real time. I have a feeling that the David Laser Scanner guys are going to make heavy use of this tech.
 
Excuse me if I am overlooking something but I do not understand how this is different than say a webcam. Is this supposed to be an innovation that has some use in the future, or just some guy hacking a kinect?
 
Excuse me if I am overlooking something but I do not understand how this is different than say a webcam. Is this supposed to be an innovation that has some use in the future, or just some guy hacking a kinect?

He isn't rotating the camera when he is moving the image, he is moving it in the program, like you would be moving the camera in a 3d editor like maya.

it is using the rgb camera to grab the color info of the objects, and the IR grid emitter and camera to determine distances, measurements, and to actually construct the 3d object that he is rotating in the program.

This is different compared to stitching two images from cameras together, in that you could not rotate around the object in software, you would need to move the cameras.

This could map all 4 sides of a scene using four, but the light interference could become too much with 4 kinects pointing at eachother.
 
Excuse me if I am overlooking something but I do not understand how this is different than say a webcam. Is this supposed to be an innovation that has some use in the future, or just some guy hacking a kinect?

I am not incredibly familiar with the Kinect, but I believe it uses an IR rangefinder to divide the image up into a grid and assign each point a range - this lets you create a "3D" picture that can be rotated with just one camera. Look back a week or so in the news to find the video of 3D with one camera. However, there are still black "shadows" where the camera cannot see since it only has one viewpoint. This video shows how the guy filled those holes and creates an image that is closer to true 3D by using two cameras together and overlaying the 3D image from each camera in order to form a model of the room in 3D that may be rotated and viewed from different angels (run on sentence there :p ). As you can see in the video, he can "see" four of the six sides of the box as he rotates the camera.

I think this stuff has been done before for movies and such, but as far as I know this is the first time that it has been done in so simply and cheaply in real-time. Obviously there's a lot of work to do, but I would call it innovation. I actually wonder if Microsoft had any idea that people would do so much so quickly once they released the hardware that made this possible.
 
Very cool stuff. Can't wait to see what he can do when he refines his calibrations and such.
 
Excuse me if I am overlooking something but I do not understand how this is different than say a webcam. Is this supposed to be an innovation that has some use in the future, or just some guy hacking a kinect?

did you just watch 10 seconds of the video and close it or something? :rolleyes: he made a 3d representation of the scene in front of the 2 cameras
 
Seems he's making progress.

I was impressed with his previous posted video with only 1 kinect. But this is pretty nice.
 
Imagine the possibility of things like online virtual tours, you could zoom in an check ouch any specific item, think of an online museum. Or doing an online class, and being able to zoom in on the notes instead of the teacher, than having the teacher start doing a demo of something and you would be able to pan around, zoom in and out and study the demonstration to check it out in 3d, allowing you to get the full scope of it without needing to cram tons of people into a classroom.

And of course 3d porn where you can view it from the angle you prefer :D lol
 
And of course 3d porn where you can view it from the angle you prefer :D lol

I am sure there are people already constructing rooms for this. Live streaming porn where you control the camera. lol technology if this guy can get the program calibrated with 4 cameras for complete capture I am pretty sure he could make some money.
 
Shouldn't be long before the porn industry gets into this tech. Being able to choose your own point of view in a porn. Hrm...
 
Wouldn't you only need 3 Kinects, each positioned 120 degrees apart from each other, to get full 360 degree capture of an object or scene?

-Cool-
 
Think of the user controlled instant replays in sports! Would love to watch hockey or soccer like this!
 
Ok, Sony is very very wrong ... Kinect is far cooler.

Think MS has a winner here.
 
Wouldn't you only need 3 Kinects, each positioned 120 degrees apart from each other, to get full 360 degree capture of an object or scene?

-Cool-

That would get you 360 degrees on a flat plane, you would need more cameras (2?) for top/bottom etc.
 
did you just watch 10 seconds of the video and close it or something? :rolleyes: he made a 3d representation of the scene in front of the 2 cameras
If you get 2 or more web cams and position them in known places, you can do the same thing this kinect is doing. When it was ONE camera it was neat, but going to 2 or more just turns out acting like normal eye balls. The coder would then have to determine the algorythym to determine distance... Any If you were futher back it would be difference sizes and such.

All the person who decided to use 2 crappy web camera's is to know the size of one of the object in the picture. Thats what the Kinect does, and heck, we know that the Sony Move knows the size of the glowing ball at all distances through simple math, and thats one camera, so could Sony move make me put on 2 gloves with tiny lights and all of a sudden I have a Kinect interface with one Camera, or should they add 2?? One in the upper far right, one in the upper far left, some bright led lights on your gloves and you have a Kinect type interface... Yes its not gloveless, but it maybe more accurate....

I have to save I love my Kinect.... and I love my Move.... I think a hybrid system could really be kick ass...
 
Wouldn't you only need 3 Kinects, each positioned 120 degrees apart from each other, to get full 360 degree capture of an object or scene?

-Cool-

on simple objects it would work, but with weird and complex shaped objects you'd have areas none of the cameras would get
 
The coder would then have to determine the algorythym to determine distance... Any If you were futher back it would be difference sizes and such..

This is what Kinect already does for you... and it does it with an infrared camera I believe and creates a 3D representation of the depth of objects it sees...

Sony uses the ball to determine the distance and position in 3D space, and then accelerometers / gyros to calculate the direction it's being held / movement.

The Wii if I remember right used the "sensor bar" (which was 2 infrared LED's) and a small camera in the remote to detect where it was pointing based on those two lights.

I think Wii is the cheapest/simplest, Kinect is the coolest, and Move is probably the most accurate.
 
only a matter of time before they use 3-4 kinects.

I say within reason. It's cool to see whats possible combining a few Kinects, but the point is to use the technology to develop something better, rather than a farm of 10 cameras scattered around a room. We can already do that.

-Cool-
 
Back
Top