Page 1 of 1
large & small offsets in motion curves vs. camera count
Posted: Fri May 02, 2008 8:47 am
by artoo
Hi all,
i am using the basic setup with 6 cameras.
when i record the actor all works fine, and looks well on first sight.
(little spikes in the curves though,.. but these are easy to fix)
my problem looks more like this:
ok, this picture is easy.
but very often its hard to tell where the "wrong" part is.
if you see the following:
here i can not say if the left or if the right part is wrong.
even if i scrub very far in the timeline.
so my question is, if i would use a 12 camera setup instead of 6,
would that really make my curves much better ?
tyvm!
r2
Re: large & small offsets in motion curves vs. camera count
Posted: Fri May 02, 2008 5:36 pm
by yoshi
Hello,
It looks like that jump in the track is caused by auto-gap filling during trajectorization. Auto gap filling can be disabled in Trajectorize dialogbox and you can manually fill the gap, which might produce smoother result. Also if you have more cameras there will be less occlusions and gaps.
Re: large & small offsets in motion curves vs. camera c
Posted: Sun May 04, 2008 12:15 pm
by artoo
hello,
thanks for the reply.
so i did try to trajectorize again, with the fill gap option disabled.
now i do have gaps here and there.
but only very few and little ones.
the problem with these offsets is still there.. all over all curves.
with these, the limbs and especialy the spine to little one-frame movements.. like rotating a little bit.
could i maybe send you a link to a file, so you can have a look for your own ?
when i send you the projectfile you should have all camera positions, skeleton and the 2d and trajectorized data.
it would be a great help to know if i did somthing wrong.. or if i just need more cameras.
tyvm!
Re: large & small offsets in motion curves vs. camera c
Posted: Sun May 04, 2008 12:32 pm
by Jim
Hello:
Yes, please post the project file link, we will take a look at it, or you can email the link to sales at naturalpoint .com We will then forward it to Yoshi.
I think that the problem is that each marker is only being seen by a few camera rays, and when another ray is added in, the marker is seen by another camera, it can adjust the marker position as it is now averaging 3 rays instead of two, or using different rays. The jump in the data is from the change in the ray positions.
You can turn on Rays and view them, and see how different cameras see the markers as you scrub through a capture.
With more cameras, you will have more rays and more to average over, so it will be smoother.
Usually, you can solve this issue by just smoothing the data with a 12hz filter, from the Editor. That will remvoe these raw variations in the data.
Additionally, I think there will be some software upgrades that should help with this in a month or so. We are always working to get more performance and value out of the V100 camera.

Re: large & small offsets in motion curves vs. camera c
Posted: Sun May 04, 2008 12:46 pm
by artoo
here is another picture.. maybe that makes the problem more obvious.
these 3 curves are the markers from the head.
the little spikes are no problem.
they make the head shake a little bit, but can be smoothed away.
but the offset (arrow) makes the head lead sideways a bit.
an as it happens in the time of one frame, it just looks like a error.
Re: large & small offsets in motion curves vs. camera c
Posted: Mon May 05, 2008 10:41 am
by Jim
Can you please send us the PT2 file and the Skeleton you are using? We can then watch the data here and offer better advice.