Hand tracking issue

Post Reply
liao86632769
Posts: 4
Joined: Thu Feb 04, 2021 7:40 pm

Hand tracking issue

Post by liao86632769 »

Hi, I am collecting motion data using Optitrack along with the Motive application. Tracking most parts of a body is perfect; however, I fail to retrieve correct hand poses in some circumstances.
Below is how a hand being tracked perfectly.
right.png
right.png (107.82 KiB) Viewed 2830 times
And here is the failure.
I realize that the markers are literally captured by the camera, but the model fitting toward the skeleton seems not being performed well in the failure case. In my experience, this failure does not happen frequently; however, when it happens I can hardly recover from this by just moving the hand around or shaking the hand. In most of the time I have to wait until it is fixed by itself or generate a new model. I would like to know if there is any suggestion to avoid such misfitting or any way to deal with the failure efficiently and rapidly. I appreciate for any advice and suggestion. Thank you.
Last edited by liao86632769 on Sun Feb 07, 2021 8:02 am, edited 1 time in total.
liao86632769
Posts: 4
Joined: Thu Feb 04, 2021 7:40 pm

Re: Hand tracking issue

Post by liao86632769 »

Sorry for the missing picture. Below depicts the failure mentioned above.
wrong.png
wrong.png (102.51 KiB) Viewed 2830 times
artoo
Posts: 56
Joined: Fri May 02, 2008 1:39 am

Re: Hand tracking issue

Post by artoo »

it looks like the inner wrist and outer hand are swapped.
that should be easy to fix and could be ignored during the recording.

to avoid it from happening in the first place, try placing the markers in a way so the distances between them are more different.
it looks a bit like the distances currently are fairly similar, and so motive has a hard time guessing which marker is which.

greetings!
liao86632769
Posts: 4
Joined: Thu Feb 04, 2021 7:40 pm

Re: Hand tracking issue

Post by liao86632769 »

Hi artoo, thanks for the reply.

Yes, the inner wrist and outer hand are swapped. As you said, it may be fixed afterward if we already know they are swapped. However, the problem is that they are not swapped during all time. Sometimes they are swapped back to the correct position and it is time consuming for staring at the record to find where the failure occurs and making further revision. By the way, thanks for the idea of the maker placement.
Post Reply