 |
www.ssontech.com SynthEyes Camera Tracker Forum
|
| View previous topic :: View next topic |
| Author |
Message |
joconnell
Joined: 09 Aug 2005 Posts: 27
|
Posted: Mon Mar 21, 2011 7:23 am Post subject: specular fur / feathers - shots not designed for 3d tracking |
|
|
Heya Folks,
I've got a few nightmare shots at the minute where I've to attach bits on to a bird. Apparently the bird wasn't exactly cooperative so as you'd imagine it didn't really take kindly to having tracking markers stuck to it's head and so I'm going off whatever detail I can see in the feathers that stays relatively locked over the course of a few frames. His head is twitching around quite a bit so what will happen is my tracking markers are good for a few frames, he'll rotate his head a little bit and the highlights will shift making all of my previous detail disappear. New highlights appear and since they're based on the shape of the feathers they lock pretty well. My main issue is the usual thing of having little parallax in the shot and getting an object track with a path that spikes back and forward fairly heavily in depth, giving the optical effect of the tracked object getting bigger and smaller at random frames.
I can get something but it's not ideal - again the shots themselves aren't designed or shot with 3d tracking in mind so I'm not expecting an easy solve where everything just works. I've got fairly decent 2d tracks but the point cloud coming out at the end is heavily stretched in depth (no doubt due to having very little parallax / rotation of the object) so does anyone have any sneaky tricks that they use in these type of situations? Are coordinate systems or distance constraints between trackers a better way of trying to coral your points into shape?
Cheers! |
|
| Back to top |
|
 |
damian_d3
Joined: 16 Apr 2008 Posts: 61
|
Posted: Mon Mar 21, 2011 8:02 am Post subject: |
|
|
Hello,
Syntheyes is actually pretty good in these kind of things.
But its a little difficult to give you a 100% working advise without seeing the plate or the sni file.
But as far as understand you have a small bird that you have to track, and it jitters in depth.
The pointcloud is stretched into depth.
You have a camera and a object in syntheyes. (Camera and the bird)
You should do following.
1.
Fix/improve the point cloud.
Disable the camera in the solver as soon as you are happy with it.
If you know the Lens you should lock it. If not, try to figure it out and then lock it (set it to known).
Setup the rootframes (front and back frames in the solver panel) manually.
You should have at least 7 trackers active between them and they should have good parallax. (That should already improve the pointcloud).
Then setup 2 distance constraints in the within the rootframes or switch to refine and setup 2 distance constraints there.
They should "pull the pointcloud together." (for example the bird has a diameter of around 10 centimeter. You setup a distance constraint (maybe 8cm) to set the scale. Then you solve... The point cloud is streched... You pick a tracker that is far away and constraint it (maybe with 5cm) to some tracker that is closer to the center.
Turn on "Contraint" on the solver panel.
Solve. You should get a better pointcoud.
2.
Remove the jitter.
To remove the depth jitter, you have to be in refine mode. "Contraint" mode is on. Then you go to the Axis Locking window. (Its in the solver panel.. the more button)
Then you press the get button. That will store the current solve as a seed path. Now you press the distance button on the first frame..
Now you can go to the curve editor and smooth or edit the seed-distance curve.
If you now solve the jitter should be gone.
[/list]
If you still have problems, you could post the sni file somewhere.
It would help to see how the file looks like.
Cheers,
Damian |
|
| Back to top |
|
 |
joconnell
Joined: 09 Aug 2005 Posts: 27
|
Posted: Mon Mar 21, 2011 1:09 pm Post subject: |
|
|
Hi Damian,
Thanks for the words of advice - especially the parts about constraints!
My biggest problem is that there's no section of the footage where the bird turns his head much - there are no sections were it's head turns more than a few degrees, except for one part where it turns it's head down 90 degrees, but that's side on to camera so again it won't give me any parallax for a solve.
I've popped up an sni file from Syntheyes 64 2011 at http://www.joconnell.com/forums/106_001_wf_1436_main_v1_08.zip if you want to take a look - the 2d trackers seem to be generally quite good with nothing causing major problems, it's again just getting something usable through out the shot. I've added some extra 2d trackers and they seem to have calmed down the jittering, the shape of the solve itself is still a problem though.
Much appreciated for the help so far - I'd say you've got a fairly heavy list of shots to work on most of the time so no worries if you don't get a chance to have a look at this!
Cheers,
John |
|
| Back to top |
|
 |
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|