FreeTrack Forum

Hi all.
I'm new here, and I've tried FT since the very first version, till the fantastic 2.0 version.
All works very well, but 2 little problems ruins the perfection of tracking:

When I look back with the game head, and want to lean my head a little bit more left or right, I must move my head in inverted mode, so to lean in right X axis, I must move with my real head in left X axis.

Maybe it's normal and typical of 6dof (so even using trackir), but will be great to find a solution.

Another little problem, for my webcam (Philips Toucam Pro 2 PCVC840K) this webcam reach 60fps at 320x240 and 30fps at 640x480, but I don't need all this fps, but even in 320x240 she gives me only 22-24fps, only in 160x120 I obtain 30fps.

Toucam support only i420 and IYUV compressions and I'm using always the IYUV compression because i420 seems to be more blind respect IYUV.

I've downloaded the latest software and driver version for my Toucam, and all is setted up right, with an Exposure of 1/50 sec, Hi-gain, 30fps, and no automatic setup at all.
Could anyone help me?

luca @ 03/09/2007 - 16h02 a dit:

Hi all.
I'm new here, and I've tried FT since the very first version, till the fantastic 2.0 version.
All works very well, but 2 little problems ruins the perfection of tracking:

When I look back with the game head, and want to lean my head a little bit more left or right, I must move my head in inverted mode, so to lean in right X axis, I must move with my real head in left X axis.

Maybe it's normal and typical of 6dof (so even using trackir), but will be great to find a solution.

I'm glad someone noticed this, the 'inverted' motion makes sense when you consider that your body should be fixed with respect to the vehicle, so when you look over your left shoulder, you have to shift your body left to get a clear view behind your vehicle, which looks like a right movement on the screen (hence the confusion).

TrackIR does this also, but some people, like yourself, thought it felt wrong so a new mode was implemented called 'TrueView' that makes the translation directions move with respect to your current view direction, instead of being fixed with respect to the vehicle.

However, it creates problems with rotation, instead of your head rotating about it's center, it rotates about the translation origin, which is fixed with respect to the vehicle. I'm of the opinion that this feels more unnatural than having to account for the _apparent_ inverted translation.

Some games implement this type of tracking themselves, like Aces High 2.

A translation-relative-to-view mode is on the Freetrack wish list.

luca @ 03/09/2007 - 16h02 a dit:

Another little problem, for my webcam (Philips Toucam Pro 2 PCVC840K) this webcam reach 60fps at 320x240 and 30fps at 640x480, but I don't need all this fps, but even in 320x240 she gives me only 22-24fps, only in 160x120 I obtain 30fps.

Toucam support only i420 and IYUV compressions and I'm using always the IYUV compression because i420 seems to be more blind respect IYUV.

I've downloaded the latest software and driver version for my Toucam, and all is setted up right, with an Exposure of 1/50 sec, Hi-gain, 30fps, and no automatic setup at all.
Could anyone help me?

Maybe a USB bandwidth bottleneck could explain why you aren't getting the full fps for higher resolutions? I assume you have USB 2.0 ports with appropiate drivers. Your webcam might be connected to a port that is being shared with another demanding device, try different ports, avoid using hubs.

Thanx Kestrel, I'll try to change usb for my Toucam
now listen to this news!!

Strange becomes more strange:

I've installed FT2 and Aces High2 in another pc (notebook) with a different webcam(Creative) and here, when I look back/left I move my body in back\left and not in back\right as in my pc!!

the profile is the same, and even the setup for my 4 leds cap
How it's possible??  

Hello Luca
Nice to see you here  

You all must know what you guys own to Luca :
Six month ago, he bothers me so much that I open Freetrack sources again (after monthes) and ... I find out by accident why the TrackIR interface couldn't run. And you know the rest of the story.

luca @ 03/09/2007 - 21h00 a dit:

Thanx Kestrel, I'll try to change usb for my Toucam
now listen to this news!!

Strange becomes more strange:

I've installed FT2 and Aces High2 in another pc (notebook) with a different webcam(Creative) and here, when I look back/left I move my body in back\left and not in back\right as in my pc!!

the profile is the same, and even the setup for my 4 leds cap
How it's possible??  

There's an option in Aces High II called 'TrackIR Object Relative Mode,' if unchecked translation moves with respect to current view.

I've also noticed RFactor has a view-relative mode by default.

Six month ago, he bothers me so much that I open Freetrack sources again

A little bit of nagging can go a long way, eh Lucas?  :P

Kestrel @ 04/09/2007 - 04h02 a dit:

A little bit of nagging can go a long way, eh Lucas?  :P

Yes Kestrel, Genius cannot sleep when the world needs them    

btw, problem solved for this inverse movement, I' haven't seen the check on the "trackir relative mode" now it's really perfect!!

For the "fps" problem I've changed the usb and now Toucam gives me 30-48 or 60fps ....that's great!!!

For all the people attracted by the high-fps webcam, must be considered that there is no difference between 30 or 60fps, it doesn't modify the quality of the tracking even when the fps ingame are 90fps or more.... after that 60fps use a lot of cpu, so it is useless!!

I'm starting to think thaht the famous story of 120 fps of the trackir is a little fake!

luca @ 04/09/2007 - 12h35 a dit:

I'm starting to think thaht the famous story of 120 fps of the trackir is a little fake!

Yes, I can't understand why is it really interesting because the humans eyes seems can't see more than 25FPS...

Rémy

luca @ 04/09/2007 - 12h35 a dit:

I'm starting to think thaht the famous story of 120 fps of the trackir is a little fake!

Well may be not : TrackIR is using a tracking algorithm. This is quite different from Freetrack that is using a "revert perspective" and is able to compute angles and pans values from a single image.
While tracking, you must not loose your marker otherwise you won't be able to make the difference between them. I guess this may happen on fast move. Increasing the frame rate reduce the delta between each video frame and improve the computation quality.
But this is only my personnal analysis on the case  

the_target @ 04/09/2007 - 22h32 a dit:

luca @ 04/09/2007 - 12h35 a dit:

I'm starting to think thaht the famous story of 120 fps of the trackir is a little fake!

Well may be not : TrackIR is using a tracking algorithm. This is quite different from Freetrack that is using a "revert perspective" and is able to compute angles and pans values from a single image.
While tracking, you must not loose your marker otherwise you won't be able to make the difference between them. I guess this may happen on fast move. Increasing the frame rate reduce the delta between each video frame and improve the computation quality.
But this is only my personnal analysis on the case  

All the point tracking algorithms I could find while researching 3 point tracking, were per-image. TrackIR uses the same tracking system as Optitrack for which API docs are freely available in the public SDK:

"INPVector:: Update: Call the update method to calculate the current vector positions based on the given camera frame.  If the call to Update is successful, the properties of the INPVector interface will contain the new values."

People often quote the NP sales propaganda of 120fps as if it were gospel.  Consider the sensor resolution:

TIR2 60k pixels
TIR3 101k pixels 358X290   120fps  33 degrees fov  (most webcams 42 deg)
TIR4 101k pixels 358X290  (accurate to 716 x 290)  120fps 46 degrees fov

TIR4 is described as being '25% faster' than TIR3, which indicates TIR3 wasn't using all 120 frames per second; hardware specs don't mean much if they're not being used! Is TIR4 actually using all 120 frames per second?

The dedicated microcontroller chip has got to read the 358X290 CMOS sensor, filter it, find the points and then apply the tracking algorithm, all in an 8 millisecond window (120fps).  This would be a burden for a 3Ghz cpu, yet alone a microcontroller...

Yes, I can't understand why is it really interesting because the humans eyes seems can't see more than 25FPS...

25fps is actually very noticeable while panning, but Freetack uses interpolation to take care of that. The only advantage of 120fps would be tracking responsiveness with 8ms lag instead of 33ms lag (assuming every frame is used).

"Simple reaction time is the time it takes to react to stimuli. The average human's reaction time falls somewhere between 200 and 270 milliseconds, although athletes and others who train themselves can achieve reaction times approaching 150 milliseconds"
http://en.wikipedia.org/wiki/Reaction_time