FreeTrack Forum

Welcome, you're not connected. ( Log in - Register )

RSS >  Wii Remote Support for FreeTrack?, FreeTrack be made to work with the Wii remote?
vermin #1 11/01/2008 - 12h45

Class : Apprenti
Posts : 1
Registered on : 11/01/2008

Off line

FreeTrack be made to work with the Wii remote?

Quote form
The Wii remote contains a 1024x768 infrared camera with built-in hardware blob tracking of up to 4 points at 100Hz.
This significantly out performs any PC "webcam" available today.
It also contains a +/-3g 8-bit 3-axis accelerometer also operating at 100Hz and an expandsion port for even more capability.
Kestrel #2 11/01/2008 - 13h11

Webmaster (admin)
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007

Off line

The Wii sensor specs are not publicly available, Johnny Lee has quoted the API output which is interpolated data. TrackIR is 120fps @ 358X290, runs off USB power, is expensive, very responsive and runs hot. The Wii on the other hand runs on 2xAA batteries (for 30 hours while using sensor), is cheap, laggy and runs much cooler, also PixArt, the makers of the sensor, only seem to make 30fps sensors. Taking all this into consideration I would be very surprised if the Wii sensor is any more than 30fps @ 320x240; no better than a cheap webcam.

Pros:
Negligable cpu usage.
No IR filter removal required.
Sensor may be better than a 30fps webcam.
Readily available.

Cons:
More expensive than a cheap webcam.
Needs bluetooth dongle.
Batteries.
Attaching it to a monitor may be difficult.
zundap #3 22/01/2008 - 15h11

Class : Apprenti
Posts : 14
Registered on : 21/01/2008

Off line

this would be perfect to me, i have a wiimote and i would not need an extra camera with no mods to make.

PLEEEEASE make this happen :D
usr #4 23/01/2008 - 15h55

Class : Apprenti
Posts : 43
Registered on : 30/11/2007

Off line

Does the Wiimote really offer that much performance on just two a AA batteries? Not entirely impossible, considering that TrackIR is supposed to run their point tracking algorithm on an FPGA (which is neither energy efficient nor cheap, but the only somewhat-affordable solution considering the low number of tracking cameras sold by Naturalpoint). The Wiimote on the other hand is sold in huge numbers (probably more Wiimotes in a week than TrackIRs in a year), so for them a custom ASIC is much cheaper than an FPGA. Considering the numbers sold it is even rather imaginable that they have some exclusive sensor from PixArt. But then, having an "interpolated" resolution for point tracking isn't as fake as interpolation in regular cameras, since it is well possible to calculate the center point of a multipixel blob to more than pixel resolution (for some reason, doing this in Freetrackfilter.ax results in a massive loss of z travel, at least with the v1 four point tracking).

Another totally unrelated thing is that all my google searches for programming info on the point tracking abilities of the Wiimote have so far only brought up those resolution/frame rate/point number claims but never an actual interface to get the raw coordinates. And those performance claims look very much like all being copied from each other, without any reference to the original source. This makes the credibility of those claims at least questionable.

Btw: why would the Wiimote support the tracking of four points, if the "sensor bar" is just a two point IR point array? I would not be all too surprised if those claims of "4 point tracking" are not the limits of the Wiimote (which only needs two points) but limits of the base station and its bluetooth connection when talking to multiple wiimotes.
Kestrel #5 23/01/2008 - 16h45

Webmaster (admin)
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007

Off line

usr @ 23/01/2008 - 15h55 a dit:

Does the Wiimote really offer that much performance on just two a AA batteries? Not entirely impossible, considering that TrackIR is supposed to run their point tracking algorithm on an FPGA (which is neither energy efficient nor cheap, but the only somewhat-affordable solution considering the low number of tracking cameras sold by Naturalpoint). The Wiimote on the other hand is sold in huge numbers (probably more Wiimotes in a week than TrackIRs in a year), so for them a custom ASIC is much cheaper than an FPGA.



A look inside:
http://www.sparkfun.com/commerce/present.php?p=Wii-Internals

Considering the numbers sold it is even rather imaginable that they have some exclusive sensor from PixArt.



The Wii Remote PixArt sensor is exclusive and the specs are confidential.

But then, having an "interpolated" resolution for point tracking isn't as fake as interpolation in regular cameras, since it is well possible to calculate the center point of a multipixel blob to more than pixel resolution.



Sure, but the interpolated resolution (Wii)  shouldn't be directly compared with raw sensor resolution (webcam).

Another totally unrelated thing is that all my google searches for programming info on the point tracking abilities of the Wiimote have so far only brought up those resolution/frame rate/point number claims but never an actual interface to get the raw coordinates.



It uses HID.

http://wiibrew.org/index.php?title=Wiimote


And those performance claims look very much like all being copied from each other, without any reference to the original source. This makes the credibility of those claims at least questionable.



The claims are from the reverse engineering of the Wii Remote's HID reports.

Btw: why would the Wiimote support the tracking of four points, if the "sensor bar" is just a two point IR point array?



Most likely a leftover from R&D.

I would not be all too surprised if those claims of "4 point tracking" are not the limits of the Wiimote (which only needs two points) but limits of the base station and its bluetooth connection when talking to multiple wiimotes.



It's definitely a limitation of the remote, it simply doesn't report any more than four.
zundap #6 23/01/2008 - 17h22

Class : Apprenti
Posts : 14
Registered on : 21/01/2008

Off line

all i can say is that the wiimote works perfectly with one led, i use it with a program for mac osx ( remotebuddy ) to control the mouse perfectly. just check the glovepie forums and see that the wiimote works very well with the IR leds.

Maybe it is just an idea, but you could check with the guy from glovepie to understand how the wiimote works, since the glovepie works very good in windows and now supports trackir!!!!!!!!!

dunno for the support of the 4 point but if it works with 3 point, i guess it should be possible to make it work
Edited by zundap on 23/01/2008 at 22h54.
Kestrel #7 24/01/2008 - 14h52

Webmaster (admin)
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007

Off line

After experimenting with a Wii remote and doing some qualitative observations, the responsiveness is very good, the sensor really does seem to be doing 100fps (image noise is FAST, so it's not being interpolated). It's also good to see FreeTrack + System process cpu usage below 1% on a single core. Sensor noise seems higher than a normal webcam, not surprising given its size, resolution doesn't seem much different to a 320x240 webcam with sub-pixel accuracy. More smoothing is required due to the higher noise which reduces responsiveness a bit, similar to reducing the sensor frame rate but is still much more responsive than a 30fps webcam.

Posted Image
Edited by Kestrel on 25/01/2008 at 02h47.
zundap #8 25/01/2008 - 09h22

Class : Apprenti
Posts : 14
Registered on : 21/01/2008

Off line

Kestrel if you need a beta tester just say so!

Great news to see that the development is ongoin!
Weyrman #9 26/01/2008 - 10h01

Class : Apprenti
Posts : 14
Registered on : 26/01/2008

Off line

I would also like to help with this if it is still being investigated
Kestrel #10 01/02/2008 - 02h17

Webmaster (admin)
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007

Off line

It's been done, however there are other parts of FT 2.2 that need to be worked on before release.
Weyrman #11 03/02/2008 - 09h17

Class : Apprenti
Posts : 14
Registered on : 26/01/2008

Off line

This is great to hear! I am eagerly looking forwards to the next version.

I actually found out about this program and that Wii remotes could be used for input from different places on the same day, and was going to post a request as to whether it could be used here but was delighted to find that someone had already gat the ball rolling
xenio #12 09/02/2008 - 12h09

Class : Apprenti
Posts : 1
Registered on : 08/02/2008

Off line

I have a wiimote and I used it with the project of Johnny Chung Lee
http://www.cs.cmu.edu/~johnny/projects/wii/
without problem, it is very fast no latency but no game support.

I am a little confused about support of wiimote in freetrack version 2.1
the Webcams Compatibles page in francaise about wii remote says: Bientôt compatible avec FreeTrack !
the english page: FreeTrack support coming soon

Thanks.
Weyrman #13 09/02/2008 - 22h40

Class : Apprenti
Posts : 14
Registered on : 26/01/2008

Off line

Umm.....from Kestrel's post above, I would assume that support for the Wii remote has been figured out and implemented in the next release version 2.2, but that 2.2 in itself is not yet ready for release.

Patience my friend, its definitely coming!
Kestrel #14 11/03/2008 - 07h59

Webmaster (admin)
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007

Off line

Someone updated Wiibrew.org with more details of the Wii sensor; apparently the actual sensor resolution is only 128x96! It uses '8x subpixel analysis' to stretch it out to 1024x768. Pretty impressive.
usr #15 13/03/2008 - 23h57

Class : Apprenti
Posts : 43
Registered on : 30/11/2007

Off line

heh, did not expect it to be that bad, even more if you consider that the subpixel accuracy probably works in a way very similar to what i did in my dxfilter variants, preserving some bits on the far side of the floating point when calculating average. if they did not go for weighted mean (this has drawbacks, as noise is getting more influence that way) this "average rounding expansion" way of increasing resolution even depends on having enough blur in the optics to get soft blobs of a considerable number of pixels instead of sharp dots, this is not so good if you have so few pixels to start with.

otoh the wiimote is made for use in a fixed array, moving camera setup, where your main two (of four available) DOFs get full camera for pitch/yaw angle, unlike freetrack, which has to calculate the whole raneg of pitch and yaw from only a very small part of the image resolution.
Edited by usr on 13/03/2008 at 23h57.

 >  Fast reply

Message

 >  Stats

1 user(s) connected during the last 10 minutes (0 member(s) and 1 guest(s)).