CGTalk - Labelling question

simonal

01-19-2011, 10:52 AM

I am embarking on a research project with hours (!) of captured but unlabeled data from a passive optical mocap system (Optitrack). The data is at times of really bad quality, with lots of ghost markers, dropped frames and some really bad occlusions.

I tried to label the data in MotionBuilder but I get really bad result, with markers swapped everywhere in the timeline. I would like to know more of how the Motion Builder labelling process works under the hood. Is Motion builder using rigidbody or skeleton structure information in the labelling process, or is it working purely on a frame to frame proximity basis? Is there someway to configure the labeller to better find the trajectories?

Any other software out there that has a more configurable labelling process?

I am new to the field and would really appreciate some guidance!

/Simon

18th_Devil

01-21-2011, 05:16 PM

The data should be labeled with the software provided with the Optitrack system.

A simple example:

Data is captured on a Vicon system, in order to capture the data Vicon Blade would be used. Vicon Blade also labels all the marker data so the software knows what marker represents what part of the body.

That tracked data will be then brought into Motionbuilder for clean up and editing.

Hope this is the answer your looking for.

simonal

01-22-2011, 08:14 PM

Hi, and thanks for reply!

I agree your suggested pipeline would be the ideal way to go, but unfortunately it is not feasible in my case. First, the capture setup was done in a non-startard way (only upper body and head) for research purposes and not for animation, and second, the Optitrack software used (Rigid body toolkit) does not do labelling. Either I must use a commercial software or write my own labelling code.

CGTalk Moderation

01-22-2011, 08:14 PM

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.