Kinect technology thread

(The refered patent seems easier to read) http://www.wipo.int/pctdb/en/wo.jsp?wo=2007043036&IA=IL2006000335&DISPLAY=DESC

My amateurish interpretation is "it's like structured light, except they shine a changing pattern of dots instead of a grid".
Yes. It's a fixed pattern of random dots to avoid the possible confusion of a repeating pattern.

Basically, Kinect projects an image of dots into the scene and then measures the variation of the dots as indicative of the distance they have travelled. This also means :
Preferably, the system configuration is such that an average speckle feature size, Axcam, on a pixel plane of the imaging unit is at least two pixels.
...it looks like I was right regards the 3D camera's resolution. It's almost certainly a standard VGA or higher resolution CCD with the sampling consisting of 320 x 240 samples. This also means that the object surface is going to be a factor, as something that absorbs IR is going to give a false distance...

In some configurations, the adjustor may be controlled to adjust illumination intensity to partially cancel/smooth the effect of the mapped objects on the brightness level received at the imaging unit, e.g. the effects being due to differences in reflection coefficient (albedo), distance, and surface properties. Such an adjustor may receive inputs from the imaging/control unit, and the intensity distribution to be projected in the plane transversal to the optical path of illuminating light propagation.

This is something TOF cameras won't have to worry about, and perhpas explains early problems with Natal regards imaging dark people (skin or clothing). I wonder if jewellery can throw it out?Could be wearing a shiny watch is a bad idea.
 
I'm quite sure I read about it somewhere, but even if not, that seems a very likely explanation. They aren't using any timing. Failing that, they could always determine distance from the IR source as proportional to intensity, but then you have to worry about different materials reflecting different amounts of IR back. That's something a projected pattern is much more robust dealing with.

that seems to be the issue with Black because it's IR absorbent

 
that seems to be the issue with Black because it's IR absorbent

Black means a colour that absorbs visible light, things that are black don't necessarily absorb near-IR light.

From the near-IR images that I've managed to find:
- some clothing does not show up properly on near-IR cameras, the near-IR light mostly passes through it, but the light will reflect off the skin behind it. (it's possible that loose clothing of this type could cause problems if it partially reflects the light - creating a ghost/flickering image?).
- black leather seems to be very, very black for near-IR.
 

Sigh, it gets even more pathethic :cry:

Depth resolution is only 320x240! We can all forget about finger tracking, and at that resolution even determing whether fists are open/closed may be problematic - I don't think they'd be included in the 20 joints that are tracked either.

What do they mean by "able to track up to 6 people, including 2 active players", so it's only 2 player now? or 6 player for like quiz show games or that it can recognise (or "remember") 6 players who can take turns stepping in and out of play - but only 2 can play at one time?

The horizontal field of view is also quite narrow, I was expecting something closer to 90 degrees. It doesn't seem that the camera can pan either, only tilt - which was a mistake, considering the FOV.

And on top of this, they want $150 for it! For a 320x240 pixel camera!
 
The horizontal field of view is also quite narrow, I was expecting something closer to 90 degrees. It doesn't seem that the camera can pan either, only tilt - which was a mistake, considering the FOV.

At least one of the Forza demo's showed both panning and tilting. And the main press conference showed only panning.

Regards,
SB
 
Sigh, it gets even more pathethic :cry:

Depth resolution is only 320x240! We can all forget about finger tracking, and at that resolution even determing whether fists are open/closed may be problematic - I don't think they'd be included in the 20 joints that are tracked either.

What do they mean by "able to track up to 6 people, including 2 active players", so it's only 2 player now? or 6 player for like quiz show games or that it can recognise (or "remember") 6 players who can take turns stepping in and out of play - but only 2 can play at one time?

The horizontal field of view is also quite narrow, I was expecting something closer to 90 degrees. It doesn't seem that the camera can pan either, only tilt - which was a mistake, considering the FOV.

And on top of this, they want $150 for it! For a 320x240 pixel camera!

I agree...
 
With pattern based depth detection a lot of the cameras native resolution is lost in processing ... it's unlikely to be the camera resolution.
 
am I the only that's bothered by the fact that the Depth camera that they will be using for the tracking is below the FPS that the old EyeToy used for tracking?

it's 320x240 @ 30 FPS and the EyeToy was 320x240 @ 60 FPS


say if they did a game like Antigrav with Kinect wouldn't it be less responsive?
 
Well, there's enough lag going on that ~17 extra ms isn't going to make a significant difference regards responiveness. I've postulated that 30 fps may miss some fast movements, but if so, chances are games are designed around such limitations such that you hopefully wouldn't notice. Nothing in Antigrav tracks a player's movement where the player's movement is finished in a 60th of a second. Worst offenders would be things like energetic fighting games where a fast jab initiated between camera frames wouldn't be seen as a jab, but that seems a fringe case at the moment.
 
The predictive tracking of the whole body can't be compared to the old eyetoy.
 
The predictive tracking of the whole body can't be compared to the old eyetoy.

but in the kinect River Rush game and Sonic Free Rider you're playing it almost the same way as EyeToy Antigrav , it's tracking your jumps and hand movements and how you lean.
 
Soon, I think

but in the kinect River Rush game and Sonic Free Rider you're playing it almost the same way as EyeToy Antigrav , it's tracking your jumps and hand movements and how you lean.

So far they don't show any impressive games. You are right. These are all just like old PS2 Eyetoy games. I think they are still developing. Probably we will see some very soon with demos. Remember this can detect your hand location in front of your body so that is something PS2 eyetoy cannot do.
 
but in the kinect River Rush game and Sonic Free Rider you're playing it almost the same way as EyeToy Antigrav , it's tracking your jumps and hand movements and how you lean.

The dude behind fable said the secod wave of games is much more impressive. So we will most likely see demos of them in the fall I'd say. Sorta like hey here are our launch games but look at the cool stuff coming 5 months down the line
 
So far they don't show any impressive games. You are right. These are all just like old PS2 Eyetoy games. I think they are still developing. Probably we will see some very soon with demos. Remember this can detect your hand location in front of your body so that is something PS2 eyetoy cannot do.

that's why I said in a game like Antigrav / River Rush where the Depth tracking don't really matter because it's just tracking your jumps & side to side movements , I think PS-Eye might work better with them games because it can track your movements faster even on the Eyetoy with Antigrav it seem to be less of a delay when you reach your hands to the side and for the jumps than there is with the River Rush / track & field game on Kinect


The dude behind fable said the secod wave of games is much more impressive. So we will most likely see demos of them in the fall I'd say. Sorta like hey here are our launch games but look at the cool stuff coming 5 months down the line

yes that's for sure you can never really judge a system by it's launch line up because the more time devs have with it the better they can get to know it and make better games.
 
A developer quote (senior producer) saying Kinetc requires gamers to be standing...

Unfortunately for Xbox 360 owners, there are no plans to support Kinect. Girard said the stumbling point was having to play standing up while using Microsoft's rival motion-sensing tech.
 
Kinect Software Will Work with Seated Players When 'Developed with Sitting in Mind'

Kinect Software Will Work with Seated Players When 'Developed with Sitting in Mind'

Microsoft has clarified that Kinect is very capable of recognizing a user that's sitting, an aspect of the new tech that continues to be a point of confusion. "Kinect can be used while sitting when an experience is developed with sitting in mind," a representative for the company confirmed to Joystiq today.

As for previously revealed Kinect applications that are being developed with a seated user in mind, Microsoft pointed to navigating the Dashboard, along with using the ESPN, Zune and Video Kinect apps as "experiences where we expect people to be sitting."

Microsoft concluded that what it showed at E3 was "only the tip of the iceberg" and that it's going to be "natural for Kinect games to be designed to get you off the couch: dancing, running, dodging, bending and kicking." It's just good to know that we'll also have opportunities to stay seated and move as little as possible to connect to this motion-based gameplay technology.

http://www.joystiq.com/2010/07/07/kinect-software-will-work-with-seated-players/?
 
So either
1) The producer didn't know the real reasons for Kinect problems
2) The developers of RUSE aren't trying to design their Kinect solution to work with seated players.
3) That MS statement is incorrect and at least one developer tried integrating Kinect into their title to find it didn't track them when they were sitting down.

???
 
... or it's a work in progress. While it may not track "sitting games" reliably now, MS is working towards it in the (near) future -- starting with the apps they mentioned above.
 
Probably more what Patsu said. As Turn10 have already demonstrated seated gameplay to some media people.

So it could be that when the Producer (who might know nothing about programming) was shown an early prototype of a game being proposed using an early prototype of Kinect, may have assumed it wouldn't work, and thus didn't greenlight funds for the project Or when the developer team showed him a project that they were working on with no plans to include sitting, the producer mis-interpreted that as no ability to sit in Kinect games. Who knows.

And once a producer gets something in their head, it's sometimes difficult to get them to think differently.

Regards,
SB
 
So either
1) The producer didn't know the real reasons for Kinect problems
2) The developers of RUSE aren't trying to design their Kinect solution to work with seated players.
3) That MS statement is incorrect and at least one developer tried integrating Kinect into their title to find it didn't track them when they were sitting down.

???

The problem could be when you start the game sitting down, because the camera might have problems recognizing you height, it could be possible that if a game supports you to be seated it might require for you to stand up at the start of the game to let Kinect know you will seat down.

Comment from Che Chou Forza community manager.
I dunno, it just feels like Sony threw a Move party and nobody showed up. It's just so unimaginative. I didn't care for Wii motion plus. How is Move any different? As for sitting down to play the Forza preview we showed -- I don't expect you to know this of course since you weren't at E3, but you can play the game sitting down and Dan showed this in the demo booth.

So there you go, some reassurance or something.
laugh.gif
http://www.neogaf.com/forum/showpost.php?p=22030301&postcount=586
 
Last edited by a moderator:
Back
Top