Hi all,

I am currently performing various tests to measure the quality of captured motions, and I have thus far encountered two kinds of motion that I am unable to capture well, namely head orientation movements (e.g., yes-no motions with otherwise static body) and wrist orientation changes (e.g., circle motion of the wrists with otherwise static body).
Although I have been moderately successful with wrist motions, I must confess that I´ve had no luck with head motions.
This post is to identify if I´m encountering a limitation of the system, or if my issues are more related to my current camera/actor setup. So, as a first iteration, I´d like to know if someone has successfully captured such kind of movements.

To give you a general idea of my current setup, I use three cameras (located roughly North, NorthEast, NorthWest, with the actor facing North), good illumination, a light background, and the actor is light-skinned, dark-haired and wears blue jeans, black short-sleeved T-shirt over green long-sleeved T-shirt.

TIA,

Adolfo Rodríguez Tsouroukdissian, Ph. D.

Robotics engineer
PAL ROBOTICS S.L
http://www.pal-robotics.com
Tel. +34.93.414.53.47
Fax.+34.93.209.11.09

The software doesn't currently track the head or hands. I believe they are planning to add these features in a future release.

G.

_________________
Greenlaw
Artist/Partner - Little Green Dog | My Demo Reels (2013,) (2015,) (2017,) and (2019)

Watch a one minute excerpt on Vimeo now!

Head tracking is planned. Hand tracking is something for later versions of the software.

You say you HAVE been moderately successful with wrist motion? How are you doing it?

I know the Standford University team took the video and translated it into a silhouette video on their markerless mocap. Perhaps that might be the ticket for achieving the best mocap translation.

Performance Capture from Sparse Multi-view Video

Good luck...

Thank you all for your answers concerning current capabilities and future plans.



Wrist motions only, without arm motion: no success. Tried once or twice, then dropped the efforts.

In a motion where the arms were moving and the hands experienced small motions, I had the said moderate success. However, what I did was highly manual. I would manually set the hand pose after an initial tracking, then propagate forward, and whenever the hand pose diverged substantially from the video, I would stop the propagation, readjust manually, and resume. Lather, rinse, repeat ;)

Best,

Adolfo