> Freetrack frequent crash with overmovement of head
|
|
|
babasior |
|
Class : Webmaster (admin)
Posts : 930
Registered on : 09/07/2007
Off line Www
|
Thanks "Zagupi" for feedback !
@Zagupi :
can you give us, your freetrack.ini [center] values before crash (when normal operated) ...
I'm interresting about Z axis (and others axis too ...)
Edited by babasior on 05/02/2008 at 15h20. |
|
Zagupi |
|
Class : Apprenti
Posts : 2
Registered on : 04/02/2008
Off line
|
Sure.
Stats after perfectly normal operation, no crash:
[center]
Yaw=150911
Pitch=-88042
Roll=-1122
X=126071
Y=330810
Z=2073049
|
|
Kestrel |
|
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007
Off line
|
Thanks for the feedback Zagudi, unfortunately the state information in the ini is not very useful for tracking down an internal state problem (divide by zero).
The NAN bug is very uncommon for the clip and both the clip and the cap use
mostly the same code.
----------------
From babasior : Excuse me kestrel i have made a modification of your message (here)
So i was a mistake ... Sorry for that ... (i wanted to use "citer" and not "modifier")
So look my answer ... very sorry again ....
Edited by babasior on 06/02/2008 at 15h05. |
|
mehmetbeyoglu |
|
Class : Apprenti
Posts : 21
Registered on : 24/01/2008
Off line
|
These are my ini values at the lock-out: The skull freezed.
I am using a 3 points cap with identical dimensions as sampled in the free-track software.
PS: I didn't see any NAN error message anywhere. Where to look?
[Camera]
Fps=30
Width=320
Height=240
Name=Logitech QuickCam for Notebooks
Compressor={E436EB7D-524F-11CE-9F53-0020AF0BA770}
FreetrackFilter=SSE2
[ForceCamProp]
CameraControl=0
CameraControlExposure=-7
PropertyPage=0
0PanelState=0
0Page=0
0Control=0
0Force=0
1PanelState=0
1Page=0
1Control=0
1Force=100
2PanelState=0
2Page=0
2Control=0
2Force=100
[Defaults]
Tracking_Method=2
Threshold=180
Profile=C:\Program Files\FreeTrack\Profiles\Default.fgp
LeftClipMount=0
NumberInterpolationSamples=4
AutoNumberInterpolationSamples=1
AutoCentering=0
AutoCenteringSpeed=25
InterpWebcamFPS=30
[BinaryOptions]
LaunchatStartup=0
StartMinimized=0
ConfirmClose=1
AutoMinimize=1
AutoLoadProfile=1
AutoSaveProfile=0
[HeadAids]
WhiteSky=1
Sphere=1
Axes=1
Labels=1
Tracking=1
LimitBox=1
LimitBoxAbs=1
LimitRotation=1
[Threepoint_clip_dimensions]
x1=60
y1=60
x2=80
y2=80
[Threepoint_cap_dimensions]
x=70
y=80
z=100
[Fourpoint_dimensions]
x1=70
y1=80
z1=100
y2=20
z2=40
[center]
Yaw=159371
Pitch=-95826
Roll=-599
X=-39687
Y=485843
Z=1695208
[Offset3Pts]
YX=100
XX=53
XY=58
XZ=190
[Inverted]
Yaw=0
Pitch=0
Roll=0
X=0
Y=0
Z=0
[Sensitivity]
Yaw=100
Pitch=100
Roll=100
X=100
Y=100
Z=100
[Deadzone]
Yaw=100
Pitch=100
Roll=100
X=100
Y=100
Z=100
Zooming=1
[Games]
48091=Default
5118=Default
62451=Default
50146=Default
29678=Default
13309=Default
22490=Default
16849=Default
6096=Default
38885=Default
19441=Default
21450=Default
21475=Default
32505=Default
38138=Default
|
|
Kestrel |
|
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007
Off line
|
It's actually being caused by the square root of a negative number, corresponding with a head orientation that has somehow jumped into another dimension (complex plane). I assume this is the fault of the tracking algorithm and not your head. 
The fix will be in the next release. |
|
babasior |
|
Class : Webmaster (admin)
Posts : 930
Registered on : 09/07/2007
Off line Www
|
from Kestrel
Thanks for the feedback Zagudi, unfortunately the state information in the ini is not very useful for tracking down an internal state problem (divide by zero).
The NAN bug is very uncommon for the clip and both the clip and the cap use
mostly the same code.
So Kestrel, wich type of informations you need to help you to solved code problem ?
Do you think, FreeTrack 2.1 release with debbug tab datas can help you ?
Kestrel, you tell that "the clip and both the clip and the cap use
mostly the same code", so i think remember that i have this kind of problem with old 4 dots cap tracking method too ...
Is this bug was here in FreeTrack 2.0 release ? So the question can be :
Witch part of "tracking" code has changed between freetrack 2.0 and 2.1
pose.pas ?
Edited by babasior on 06/02/2008 at 14h57. |
|
RomD |
|
Class : Apprenti
Posts : 23
Registered on : 04/01/2008
Off line
|
Head stuck problem (rotation is always -157,51 and S is NAN when it happens):
Strange problem with 640x480 mode (LEDs aren't tracked near the right border):
 |
|
Kestrel |
|
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007
Off line
|
So Kestrel, wich type of informations you need to help you to solved code problem ?
Do you think, FreeTrack 2.1 release with debbug tab datas can help you ?
As I mentioned above I've already found the cause of the NAN bug and resolved it.
Kestrel, you tell that "the clip and both the clip and the cap use
mostly the same code", so i think remember that i have this kind of problem with old 4 dots cap tracking method too ...
The 4 point code is completely different to the 3 point code and doesn't involve any potential square roots of negative numbers so cannot suffer from this bug.
Is this bug was here in FreeTrack 2.0 release ? So the question can be :
Witch part of "tracking" code has changed between freetrack 2.0 and 2.1
pose.pas ?
It was in 2.0.
Head stuck problem (rotation is always -157,51 and S is NAN when it happens)
The NAN bug can affect both the orientation and the smoothing.
Strange problem with 640x480 mode (LEDs aren't tracked near the right border)
Oops. The internal video limits were 512. Thanks for pointing this out, most people don't use 640x480 due to the cpu overhead.
|
|
Kestrel |
|
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007
Off line
|
In the meantime... as long as ONLY the led points are tracked (no multiple crosses per led and absolutely no other light) and the model dimensions are suitable, the setup is completely stable and a NAN error cannot occur. Make sure you set the depth dimension according to what it LOOKS like in the video when you turn your head (compared with the other points) rather than what it actually is. Lastly, try and make your model deeper, with the middle point much further back than the distance between the two front points (equally, move the two front points closer together).
|
|
Stryder786 |
|
Class : Apprenti
Posts : 3
Registered on : 06/01/2008
Off line
|
Yeah, lately I have been having a lot of trouble with FreeTrack while playing Open Falcon. Sometimes it's a NAN error, sometimes it's not. It just seems to randomly crash on me, and completely locks down my game. It seems to happen within the first 30 minutes of the game, or sometimes in the first 5 minutes.
|
|
RomD |
|
Class : Apprenti
Posts : 23
Registered on : 04/01/2008
Off line
|
I noticed that when I look down slightly and rotate my head to the very right the FreeTrack head flips and is upside down. When i look up a bit it works perfectly. Rotating to the left side doesn't cause any problems at all. Is this related to the mentioned bug?
|
|
Kestrel |
|
Class : Webmaster (admin)
Posts : 780
Registered on : 13/07/2007
Off line
|
The software FAQ has been updated:
I'm using 3 point cap tracking and sometimes tracking fails for some axes and 'NAN' appears in the data readout.
The front points cannot be allowed to move behind the middle point, relative to the camera plane. This could happen for downward pitching and extreme yawing. To avoid this, make your model deeper, with the middle point further back than the distance between the two front points (equally, move the two front points closer together).
Also make sure that only the tracking points are being tracked at all times and that there is only one tracking cross per point.
|
|
BlindVision |
|
Class : Apprenti
Posts : 49
Registered on : 13/02/2008
Off line Www
|
Kestrel @ 06/02/2008 - 15h41 a dit: It's actually being caused by the square root of a negative number, corresponding with a head orientation that has somehow jumped into another dimension (complex plane). I assume this is the fault of the tracking algorithm and not your head.
The fix will be in the next release.
You are the man Kestel
Thanks -looking forward to the release of ? v2,2 then.
|