forums.naturalpoint.com

I am looking at some non-trajectorized data and the skeleton is all over the place. sometimes it "catches" the data and tracks correctly, other times it doesn't. Whats strange is that it is not consistent, some times it tracks correctly, sometimes not.

What's even stranger is that when we where capturing the data the skeleton was tracking great, no red body parts, etc. The next day I go to edit the data and I have a guy with his foot where his head should be and his arms doing things that I think are illegal in several states.

but I digress.

When I go to trajectorize the data i am left with an unusable take. I can look at the original point cloud data and my monkey brain can connect the dots; even if the software is having trouble.

My natural instinct would be to look for some way to tell the software "HEY YOU this is the foot marker RIGHT HERE" and help the software do the trajectorizing. Alas that feature seems to be absent...

I am using the right skeleton/data combo as well, so that's not the issue. Any help would be appreciated. I can send some data to you as well.

thanks much.

Could you send the files to support?

sure thing, i just sent them.
thanks!

OK,

We are looking a little closer at the data, but initial thoughts��.

1 � We recommend starting and ending a take in a t-pose. This allows the software to better match up the skeleton to the tracking points. It also allows a skeleton to be recreated for that particular take in case the markers and suit have moved over time.

2 � Try adjusting the markers � making sure to look at the neck, shoulder and hip markers.

3 � Try adjusting the height of the skeleton itself. The current skeleton looks too tall. The head markers should be closer to the top of the head, instead of in the middle of the head itself.

One more tip, you can reset the skeleton on the fly by pressing Ctrl-r when the root hip bone is selected and flashing. Use the Select tool to highlight the root hip bone.

The reason it is always different is the solver is trying to find the best solution and it can start in different forms, so as Vince suggested, it is best to start and end in the T-Pose, as this helps the solver a lot. When you started your 2D take, the skeleton was already being tracked nicely, probably from a T-Pose, and when you hit "record" only that data on gets saved, so when you trajectorize, the skeleton solver is starting from that point, not having any previous knowledge of the skeleton pose.

Let me know if Ctrl-r works for you, you can hit it several times in a row to get it to "catch".