forums.naturalpoint.com

Hey all,

Just checking in to see if any OptiTrack users have successfully created a facial mocap system using either one or three cameras. We're looking into doing the same; any tips would be appreciated.

We're looking at two options:

1. single camera attached to head rig. Project the 2d marker positions onto face model for pseudo-3d marker animation on model.

2. three cameras fixed to floor. Somehow merge tracked marker positions from the three cameras to a 3d point positions (xyz).

We definitely would opt for number two, but it seems that OptiTrack SDK is not able to merge three cameras' data in order to output xyz marker data. If I understand correctly, the SDK doesn't even track markers currently - it outputs them in a stream, and the application using the SDK needs to sort them out per-frame.

Has anyone attempted to create this kind of a system? We're contemplating waiting for the next version of the SDK to be released, which may or may not include multi-camera marker tracking. Until then, it seems that the only feasible way to go would be the single camera 2d method, which will yield a less believable final animated face.

At any rate, I just thought I'd post this to the community, as there is no other thread talking about it yet. Thanks in advance to those who respond!

Have a good one,

Erik
EGirl Interactive LLC

Erik:

You have pointed out an ideal application for OptiTrack, one that we know some users are working on, although we are not sure if that info is meant to be public.

You are correct, the current SDK does not corrolate the marker positions from mutliple cameras, but we may do this in the future. The host applicaiton must track the markers and derive the 3D position.

We do have mulitple users who have made multi-camera systems, we know because we helped fix various bugs they found! It is doable, but would require quite a bit of work.

Currently OptiTrack is meant to be a development platform, and that works well for some users, in the future, we will make it more of a "complete" product, based on feedback gathered from this forum and users.

Hey, if you know of any great programmers out there who want to come and work on OptiTrack applications we are hiring! I might just make that a seperate post now that I think about it.

quote:Originally posted by egirl:
Hey all,

Just checking in to see if any OptiTrack users have successfully created a facial mocap system using either one or three cameras. We're looking into doing the same; any tips would be appreciated.

We're looking at two options:

1. single camera attached to head rig. Project the 2d marker positions onto face model for pseudo-3d marker animation on model.

2. three cameras fixed to floor. Somehow merge tracked marker positions from the three cameras to a 3d point positions (xyz).

We definitely would opt for number two, but it seems that OptiTrack SDK is not able to merge three cameras' data in order to output xyz marker data. If I understand correctly, the SDK doesn't even track markers currently - it outputs them in a stream, and the application using the SDK needs to sort them out per-frame.

Has anyone attempted to create this kind of a system? We're contemplating waiting for the next version of the SDK to be released, which may or may not include multi-camera marker tracking. Until then, it seems that the only feasible way to go would be the single camera 2d method, which will yield a less believable final animated face.

At any rate, I just thought I'd post this to the community, as there is no other thread talking about it yet. Thanks in advance to those who respond!

Have a good one,

Erik
EGirl Interactive LLCTake a look at http://forums.naturalpoint.com/cgi-bin/ ... 7;t=000035

He talks about camera calibration and finding the position of markers in 3D space. When I have more time, I'm going to work on this.