1

Maybe I missed something.  Can you set an angle to marker points so that the angle follows the trajectory points.  This way the angle could be viewed throughout the video. (i.e. knee angle throughout gait video)

Thanks,

Zach

2

Hi,
No, it is not currently possible.

It was necessary to improve the tracking performance and robustness before implementing multi-point tracking.
It is planned as a future improvement though.

3

zem3 wrote:

Maybe I missed something.  Can you set an angle to marker points so that the angle follows the trajectory points.  This way the angle could be viewed throughout the video. (i.e. knee angle throughout gait video)

Thanks,

Zach

Has this feature been revisited recently?

4

Well, it is in the pool of new features that will be scheduled for integration during the next series of experimental versions.

5 (edited by coachgeo 2011-06-13 15:52:51)

zem3 wrote:

... (i.e. knee angle throughout gait video)...

uewwwww... me likey.  Look forward to improvement of track feature that will open up doors to do more things such as the one wished for here.


Would it help the Auto track work better if the user had the ability to mark/Select the object desired to be tracked in as many frames/Key Images manually as they desire to improve the auto tracks ability to follow that object. Would assume the more manual marks made, the more accurate "auto" tracking becomes.  Guess this would be maunally assisted auto tracking.?

I was tracking the hips of a Artistic Gymnast on High Bar with objects in the back ground and fore ground. Those objects the auto tracker would missmatch with the one I had selected on the athlete and cause major missdirection of path.  Miss matches were with things like wall trusses, cables that hold up the bar, verticle post that holds up the bar, support cables on nearby  equipment etc. 

Now that I think about it.... another Possible solution; instead or in addition too, idea above of multiple selecting of object to be tracked ..... how about a tool with adjustable size (width/hight)*  that you can sweep around a frame(s)/Key Image(s) "deselecting" large amounts of data in areas of frame(s) so that the auto tracker engine does not even looks at it data in those areas. 

My feeble brain seems to think each of these should allow the auto track engine to be even more accurate, even more so with both used in conjunction to one another cause it ends up giving the auto tracker engine a more finite set of data to work with

*adjustable size similar to how you can adjust the size of a paint brush in photoshop/drawing programs.

6

coachgeo wrote:

Would it help the Auto track work better if the user had the ability to mark/Select the object desired to be tracked in as many frames manually as they desire to improve the auto tracks ability to follow that object. Would assume the more manual marks made, the more accurate "auto" tracking becomes.  Guess this would be maunally assisted auto tracking.?

...my feeble brain suggests that that should help too - even just a start point to end point could create a likely corridor so that auto tracking could look for something in between the shape/light/colour of A and Z between A and Z and probably on a line between them, a mid point or bounce/impact point or multi manual points should therefore help even more...
Nice idea

7 (edited by coachgeo 2011-06-16 02:50:37)

The deselecting and selecting data points in frames/Key Images idea presented above really is hitting me as to making huge differnece in the ability of this program's tracking engine.  For example... in the real world of Biomechanical analysis.... the researcher who is using video... goes thru alot of preperation to make it easy for the camera to pick up undistracted data and send it to the computer for processing.  Soooo... they use  rooms void of background distraction..... they place refective markers on joints of the athletes body and or object. They;  prior to data collection, sych the computer program with data collection points so during collection it will deal with ONLY those points.   

example see http://www.youtube.com/watch?v=nnUQh_ig0kY


What I am proposing is to do this all on the computer instead of physically.  Mark/Select data collection points on the body/object (same as we do now for tracking.. but allow for multi frame selection/Key Image for identification of same point on moving person/object thru out the vid) and wipeing out distracting data with a Deselecting sweeper

infact with select and deselect tool along with the software's tracking engine IMHO it might even be possible to convert to simple 3d stick figures like used in the sample video.  WOW... talk about jumping the potential of this program sky high.

Joan... hugs and kiss, lots of beer, anything I can give you to encourage you to dive into this area.  Wish to hell I could give ya money.... -blush sad  starving coach here.  Literally- filing for bankruptcy this week  sad



PS---- THANK YOU SO MUCH TO ALL OF YOU WHO HAVE BEEN PART OF CREATING KINOVEA