Overview of a motion capture session. One person is performing, guided by a technician while another one is checking the real time tracking quality for possible markers issues.
During my time at Miralab, University of Geneva, I used to be responsible of the Motion capture laboratory. The tasks were various but always to make human motion capture and deliver a dataset of point in space or a corresponding animation. I also collected an extensive dataset from salsa dance motion for my PhD. I will describe here the procedure.
The first step is to suit up our human subject with a pyjama-like grey suit. Then, 52 marker are added using Velcro all over the body. These markers have to be put in very specific places corresponding to anatomical “clues”, such as joints or head. Then a calibration motion is required, that will “teach” the software how the body is made (limb length etc..).
Then the recording can take place. After recording, we can run the calibration motion through the solver that will animate a predefined avatar body on top of the moving skeleton. This step can be done also afterwards.
Many difficulties came along the procedure, such as calibration motion quality and occlusions that were very tough to deal with. In the example of salsa dance, both people had to wear 52 markers and some of which could be easily occluded when in the embrace position.
Illustration of where the markers should be placed on the body to have an efficient motion capture..
Motion capture also includes face motion capture. For this request, we place small markers on the face of the actor, surrounded by cameras. Then a specific configuration is setup and allows to map the results onto a virtual avatar.
In the video is an example of salsa dance motion capture. We can see that sometimes the tracking is having difficulties due to the occlusion when moving the arms around. For the information, this video has been published on social network and has reached 800,000+ views, although stolen by some people. I intervened to take the credit back but it was too late. Now what you can see here is the original, and I took the precaution to add a watermark this time 😉
Tools
Vicon motion capture system – Blade 2 – Visage Capture – Motion Builder
Methodology
Motion capture pipeline: makers placement, calibration, rigging and animation.
Outcome
Dataset of motion capture (3D points in space + time) at 120fps.
Animated characters usable for scene or in video games.
Facial animation and/or health measurements