Change font size
It is currently Tue Jul 28, 2026 12:48 am


Post a new topicPost a reply Page 1 of 1   [ 8 posts ]
Author Message
PostPosted: Tue Nov 25, 2014 1:54 pm 

Joined: Tue Oct 21, 2014 4:02 pm
Posts: 5
This has got to be the number one time killer when tracking my mocap data. The shoulders ALWAYS try to turn inward while solving! I have to babysit the tracking progress and pause every few frames to re-rotate the shoulder to a natural position. It's very frustrating.

Has anyone else encountered this issue? Is there a way I can tell the software that "hey, it isn't natural to touch the back of your hand to your waist"?!!!


Top
 Profile  
 
PostPosted: Wed Nov 26, 2014 5:44 am 
iPi Soft

Joined: Wed Jan 11, 2012 6:12 am
Posts: 2355
Location: Moscow, Russia
Is this color or depth video?
I suppose it's color, and you just need to adjust actor colors and/or scene lighting better. Or make the lighting in mocap area more ambient in case there are harsh shadows in the video.

Please share your project and video to get more specific recommendations. You can use file sharing services for big files.


Top
 Profile  
 
PostPosted: Wed Nov 26, 2014 2:33 pm 

Joined: Tue Oct 21, 2014 4:02 pm
Posts: 5
Thanks for the quick reply!

This problem is happening when tracking from the depth video (I have not tried tracking color video yet). The shoulders just seem to default to an inward rotation which always results in bad tracking that constantly loses the lower arm when the elbow rotates. For our recording session we used two xBox Kinect cameras set slightly off 180° from each other (full 180° caused weird flickering) and placed drop cloths in front of the walls to minimize the no depth (yellow) data points.

I've attached a zip file for the iPi project and calibration scene for one of the shots I am currently working on, but this issue seems to happen with all of the mocap tracking of our data in iPiStudio. Please let me know if you need any other files to view the project. We recorded in 2-3 minute long takes so the original data files are pretty large (3GB+).

Any tips you can provide would be really helpful. I appreciate it!

Image
screenshot from frame 2050 of the project


Attachments:
File comment: iPi project file and calibration file for SarahDance005_tracking
SarahDance005_tracking.zip [2.43 MiB]
Downloaded 709 times
Top
 Profile  
 
PostPosted: Sun Nov 30, 2014 1:38 am 
iPi Soft

Joined: Wed Jan 11, 2012 6:12 am
Posts: 2355
Location: Moscow, Russia
Please provide the iPiVideo record for that project (using some file-sharing service like Drop Box or Google Drive).

Tip. You could use compression options in iPi Recorder to reduce record sizes (see COMPRESSION group on RECORD tab).


Top
 Profile  
 
PostPosted: Mon Dec 01, 2014 9:44 am 

Joined: Tue Oct 21, 2014 4:02 pm
Posts: 5
I've uploaded the ipiVideo to Google Drive. You can access the file here:

https://drive.google.com/file/d/0B4yncG ... sp=sharing


Top
 Profile  
 
PostPosted: Mon Dec 01, 2014 9:53 am 
iPi Soft

Joined: Wed Jan 11, 2012 6:12 am
Posts: 2355
Location: Moscow, Russia
Thanks. I'll check it and get back to you with feedback.


Top
 Profile  
 
PostPosted: Tue Dec 02, 2014 7:42 am 
iPi Soft

Joined: Wed Jan 11, 2012 6:12 am
Posts: 2355
Location: Moscow, Russia
I have looked at your video.
Unfortunately, there is no straightforward way to fight this problem. The error comes from the nature of optical tracking - it is hard to determine the rotation of arm's joints when it's almost straight. So there is a need to apply some heuristics to define what we can't see. And in this case it tends to default joint rotation (as in T-pose).
In V3, we are working to improve these heuristics to make a motion more continuous and natural. Though this work is not complete yet, I encourage you to try V3 on your projects. It already produces far less tracking errors than V2, so less need for clean up.

Also, I have a few general recommendations based on your video:
  • When tracking dancing motions, try tracking options Shoulder positions tracked from video and Very flexible spine. They may produce more desirable results.
  • When using depth sensors, wearing boots with reasonably thick sole improves the accuracy of feet tracking.
  • Wearing a bit tighter pants may improve tracking. Such pants as in this video hide minor movements in hip and knee joints.


Top
 Profile  
 
PostPosted: Tue Dec 02, 2014 10:12 am 

Joined: Tue Oct 21, 2014 4:02 pm
Posts: 5
I had thought that the shoulder issue might be inherent in the way the software tracks the depth information, it's just challenging because it always seems to prefer rotating the shoulders inward.

Thank you for looking into this, vmaslov. I really appreciate all of your help! Those extra tips will definitely help for our next recording session as well. I'm also excited to hear that you are working on a V3 and I would definitely like to try it out. Do you know where I can find a download for this version?

EDIT: I just found the download for version 3. Here's the link for anyone else:

http://docs.ipisoft.com/iPi_Mocap_Studi ... _3.0.2.171


Top
 Profile  
 
Display posts from previous:  Sort by  
Post a new topicPost a reply Page 1 of 1   [ 8 posts ]


Who is online

Users browsing this forum: No registered users and 44 guests


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 post attachments in this forum

Search for:
Jump to:  


Powered by phpBB® Forum Software © phpBB Group
610nm Style by Daniel St. Jules of Gamexe.net