FreeTrack Forum
FreeTrack Forum > FreeTrack : English Forum > Support : Tracking System > Difference between models
luca | #1 05/02/2008 - 03h03 |
Class : Habitué Off line |
In some older posts here, someone tolds about the obsolescence of 4 leds cap models (the first model for FT).
At the moment I've built a 4 leds cap (long time ago) and a 3 leds clip (recently) but there is a great difference. 4ledsCap...even with a bad settings is still the best for stability and precision 3ledsClip...even with a perfect settings, gives many unwanted movements and great stability problems So, I asked myself about 3leds clip....Is this new model a real revolution or a little brother of 4ledsCap? My clip is 18 30 60 40 maybe is too small?how to obtain the 4leds stability with the comfort of the clip?
Stay FREE
|
Kestrel | #2 06/02/2008 - 13h07 |
Off line |
I'm a clip user, I believe it to be the best of the tracking methods atm because it has one less led to power than 4 point, simpler model construction, convenient attachment to headphones, the rotation axes are clearly defined, it doesn't suffer from NAN bugs (or very rarely) and parasitic translation can be significantly reduced. Note that the 3 point models are more sensitive than 4 point due to differing internal parameters, this is easily accounted for by reducing sensitivity.
|
luca | #3 07/02/2008 - 11h27 |
Class : Habitué Off line |
I agree with you Kestrel, the clip is much more confortable for many reasons(all the reasons you've described) but...is very sensitive and I cannot reduce sensitivity parameters because lowering sensitivity needs a greater head rotation.
At the moment I turn my head 40° for Yaw and Pitch, and my natural heart pumping produce translations in the game So, following your suggestions, the question is: How to reduce sensitivity with the same head's rotation ratio? maybe too small? (values in mm)
Stay FREE
|
luca | #4 14/02/2008 - 12h04 |
Class : Habitué Off line |
almost solved, but not at all
Even in 2dof games, using a 6dof profile (no axis unchecked) gives a better stability,.... but little unwanted movemet caused by 3 leds clip's natural high sensitivity is an unsolved problem. So..if you use 2dof games 1led or 4led cap gives best results Someone knows if this is a common problem even in TIR trackclip?
Stay FREE
|
Kestrel | #5 14/02/2008 - 13h26 |
Off line |
I can't understand why you can't just reduce sensitivity...
The four point cap has a very small internal sensitivity multiplier which will be increased to make it equal to 3 point clip and cap sensitivity in the next version. |
luca | #6 20/02/2008 - 11h55 |
Class : Habitué Off line |
It's simple...if I reduce sensitivity I must turn my head to 180° to obtain a 45° gamehead rotation. My target is to rotate@45° my real head to look at my back. In a few word, the sensitivity factor changes not only the sensitivity but also the rotation space.
Stay FREE
|
Kestrel | #7 20/02/2008 - 13h23 |
Off line |
By definition, if both the 3 point cap and 4 point cap are mapped to the same rotation space, then they both have the same sensitivity. If you're speaking about unwanted movements caused by camera noise, in my qualitative experiments there doesn't appear to be any difference between 3p and 4p in this regard. |
luca | #8 20/02/2008 - 18h36 |
Class : Habitué Off line |
yes kestrel, for the cap it's all ok...
but the problems are only for the clip (3 points clip) and this is the model I've built for me. I've got even a 4 points cap and it's all ok, but I hate to wear a cap during the game, the clip is much more comfortable but 3points clip's sensitivity is not the same as for the 3 or 4points cap
Stay FREE
|
Kestrel | #9 21/02/2008 - 02h02 |
Off line |
If <insert tracking methods here> are mapped to the same rotation space, then by definition they have the same sensitivity. According to Wiktionary, 'sensitivity' is "the degree of response of an instrument to a change in an input signal." Can you please clarify, in as many words as possible, exactly what you mean by 'sensitivity' because what you have described cannot possibly relate to sensitivity. I suspect what you really mean is jittery, jumpy, shaky, noisy tracking, most likely due to your 3p clip model being too small in the camera frame.
|
luca | #10 22/02/2008 - 02h35 |
Class : Habitué Off line |
Sensitivity for me is the response to an input signal.
For high sensitivity, I mean a translation for every minimum movement of my body For low sensitivity I mean the production of translation only in the case of large movements of my body This parameter setted to 50 for yaw and pitch produces too many unwanted skull movements. So, when I want to see a point, the view is not stable but as you said, jumpy, jittery and shaky, caused only by my Heartbeat or natural little movements of my body....in this case I say that the view is too much sensitive. At this point, I need to reduce the sensitivity parameter to erase all the unwanted movements caused by my body, But I cannot do it, because sensitivity parameter change also the degree of movement. I don't think that my 3p clip is too small in the camera frame, is 90mm long and I'm 75-80cm from my webcam. Consider that with the same webcam settings and a different profile, the 4p cap model is 2 times more stable! btw, here you can see a video made with 3p clip, as u can see I cannot be fix to a point
Edited by luca on 22/02/2008 at 03h12.
Stay FREE
|
Kestrel | #11 23/02/2008 - 02h35 |
Off line |
Increase smoothing.
In 2.1, 3p and 4p cannot be directly compared, they have different sensitivity multipliers which also affects the smoothing effectiveness. In 2.2, they have the same multipliers, allowing them to be directly compared; I assure you there is no significant quality difference between them. |
luca | #12 23/02/2008 - 14h48 |
Class : Habitué Off line |
Increasing the smoothing does not help for this work because when the FT-skull operates Unintentional movements, with smooth these movements are only a little bit slower and more difficult to correct, due to a lower real head response.
That's what I mean: without smoothing with smoothing
Stay FREE
|
Kestrel | #13 23/02/2008 - 16h03 |
Off line |
The 'smoothing' control in FT is actually a dynamic deadzone to make the view stable for small movements and/or image noise.
|
luca | #14 24/02/2008 - 15h28 |
Class : Habitué Off line |
Surely adding a great amount of Smooth, the view becomes more stable, but at the same time, slow and not very responsive.
Basically, you are telling me that with the 3p clips model you cannot reach the tracking quality of 4p cap? This is a video showing 4points cap in action
Edited by luca on 24/02/2008 at 16h30.
Stay FREE
|
Kestrel | #15 25/02/2008 - 06h51 |
Off line |
Just to summarise, to make things absolutely clear:
* For the third and final time; all things being equal, 3p and 4p both exhibit noise to largely same degree (the information from the extra point in 4p is wasted in the algorithm). * 3p has a larger internal sensitivity multiplier than 4p in 2.1. * The difference in internal sensitivity can be accounted for in 2.1 when going from 4p to 3p by reducing sensitivity and increasing smoothing. * There are values of sensitivity and smoothing in 2.1 that make 3p and 4p tracking quality match. * 4p sensitivity (and consequently smoothing) will be made equivalent to 3p in 2.2 |
FreeTrack Forum > FreeTrack : English Forum > Support : Tracking System > Difference between models
> Stats
1 user(s) connected during the last 10 minutes (0 member(s) and 1 guest(s)).
Powered by Connectix Boards 0.8.4 © 2005-2024 (8 queries, 0.020 sec)