Kinect technology thread
- Thread starter LightHeaven
- Start date
grandmaster
Veteran
I'm not sure about the six player bit (ie, I don't know, not that I think that is wrong) but most of that is accurate. The low resolution of the depth map is what I find most surprising.
it's not 6 players only 2 people can be active meaning the other people are just there they can't play at the same time.
Yet another downgrade?
Whats that 7 or 8 already, it will be amusing to see a timeline of how the hardware has evolved over time
At this rate what will the final specs be when it actually launches?
1 joint and 1 person
Whats that 7 or 8 already, it will be amusing to see a timeline of how the hardware has evolved over time
http://www.eurogamer.net/articles/digitalfoundry-primesense-article?page=2
At this rate what will the final specs be when it actually launches?
1 joint and 1 person
I wonder if it has less to do with the actual camera than consuming less CPU and memory resources when processing data from Kinect.
Edit:
640 x 480 x 32 = 9 830 400 bits
320 x 240 x 16 = 1 228 800 bits
Having lower resolution undoubtedly lowers the amount of data processing they need to perform. How much is really not possible to know, but that memory saving is pretty big. I imagine Kinect would need to store a buffer of several frames for some API functions, so memory could be consumed very quickly. Or maybe it wouldn't, and it would only have to save the information relevant to the skeletons, which should be considerably less data.
Lowering the processing requirements might reduce latency. Who knows?
Edit:
640 x 480 x 32 = 9 830 400 bits
320 x 240 x 16 = 1 228 800 bits
Having lower resolution undoubtedly lowers the amount of data processing they need to perform. How much is really not possible to know, but that memory saving is pretty big. I imagine Kinect would need to store a buffer of several frames for some API functions, so memory could be consumed very quickly. Or maybe it wouldn't, and it would only have to save the information relevant to the skeletons, which should be considerably less data.
Lowering the processing requirements might reduce latency. Who knows?
Last edited by a moderator:
just like the videos from E3 the other people was just dancing for the hell of it.
and with that game you're not controlling the dancers you just match the moves on the side as they go by.
It would make backward compatibilty on the next xbox more difficult.
Is there anything more than rumors that tell us it doesn't have any processors? I don't think anyone has leaked internal shots yet, have they?
It doesn't work like that. It emits IR light, and then it measures reflected light to create distance data for each pixel. Essentially you create a sort of 3D depth map. Obviously you don't have a full 360 degree view of the subject. They use that 3D depth data, and process it to recognize human forms. The players human form is then tracked as a skeletal body, which is updated for each frame captured by the camera. The camera itself is not incredibly sophisticated. It's the API behind it that took a lot of time and research.
corduroygt
Banned
How can a camera capture color in 32 bits? This is not 3D rendering where you have an alpha channel, cameras are either 24 or 36 bits in my experience.
why ? just have the waternoose cpu in the kinect do the work. Why would an xbox 3 using a kinect not be able to run old games ? The waternoose cpu would still be in the kinect. Just make drivers for it to work on the xbox 3. If they want to do kinect 2 just use a newer watenoose if the waternoose isn't enough
Get your technology to market and we'll compare them.
I was referring to the difference for a frame of information for the time-of-flight camera. I believe the original specs listed were 640x480 with 32 bit precision at 30 Hz. I could be wrong. It may have been 640x480 with 16bit precision at 30Hz. In that case it would be 4 915 200 bits per frame, which is still significantly larger than the reduced resolution that may now be used. I'm just curious to know if they've chosen to limit the resolution because of the impact on memory and CPU, rather than a limitation in the tech.
This is assuming those specs are correct.
Well we'll see. Tracking a skeleton with a time-of-flight camera seems far more logical to me than doing it with heavy image processing, where the amount of guess work would be even greater. Maybe your way can be done, but it seems like it would be far more prone to error and far less efficient. Obviously I'm not an expert, but this camera tech was invented for a reason, before Microsoft ever got their hands on it.
I'm sure it's not perfect just like kinect isn't perfect. There are errors either way and as far as I gather it's doing some pretty decent image processing to get the skeleton out of the images anyways. It's just it seems like an evolutionary step from webcams more than a revolutionary one.
Silent_Buddha
Legend
Sure but then have a 6 foot tall person stand next to a 5 foot tall person. Despite the fact that they are right next to each other a 2D camera using size as a measure of distance will assume the 5 foot tall person is farther away from the screen. Now toss in a 4 foot tall kid and he'll be assumed to be even farther away.
Next dim the lights. Ut oh. Now that 2D color camera is almost useless.
Have any background objects the same or similar color to what a person is wearing? Ooops, more problems.
With a camera that can actually judge distance, you could in theory have someone painted the exact same color as the background, and the system would still be able to determine where the person is and figure out all those skeletal points. Heck it should be able to track a person in total darkness.
Regards,
SB
Similar threads
