Vicon Nexus auto-labelling help

[Arick Auyang]

Does anyone have good success using the auto-label feature with the Vicon Nexus 1.7.1 motion capturing software? I am now using the new Nexus system but came from the old Vicon Workstation software. While some things are still similar, I am not having success with Nexus. The following is what we are currently doing. We collect a static trial with the subject standing in anatomical position. We use the vicon marker set and full body plug in gait model. I label all the markers. And after this, I have been using a bit of trial and error. I try to fill gaps, run static gait, create autolabel static calibration. In the dynamic trials (running indoors across a runway), when I use the auto label function in the pipeline, it just makes a mess. I've resorted to just manually labeling, but the sheer volume of data has warranted me a few minutes to post here and request help. If you have success using the standard full body vicon marker set with the plug-in gait model with auto-labeling feature during dynamic trials, I would appreciate your insight into which steps you use in the pipeline during the static and dynamic trials and any "magic" settings you have found to work well in the auto labeling or other pipeline steps. Thanks!

Arick

[Eric Walker]

Re: Vicon Nexus auto-labelling help

Based on my experience, I have found that the successful use of the reconstruct and label pipeline depends mainly upon the quality of the motion capture data. When it works well, it is great for speeding up the post processing. However, when the auto-labeling starts making mistakes (markers changing location between legs or with other markers), it ends up taking more time to break and re-label the mislabeled trajectories than manually labeling from the start. We have experiments using both the lower body and full body Plug-in gait model, and certain protocols seem to work well with the auto-labeling, while others the auto-labeling makes things worse. I think that ultimately the success of the auto-labeling depends upon the marker drop out, mainly how many markers drop out at the same time or over the same range.

One thing you could try to possibly improve the auto-labeling success is to adjust the settings of the core processor in the reconstruct and label pipeline. I would make sure to save the original pipeline provided from Vicon, and edit a new version of that pipeline. There are some settings (and more under the show advanced) that you may need to adjust to better fit the conditions of your experiment. Since you noted that the dynamic trials involve running, you could try to increase the marker movement speed and see if that improves the marker reconstruction and labeling. The best option is to make sure you are collecting the best marker data from the start: make sure that markers are not getting covered up during movement and camera placement is optimized. Adjusting the reconstruction parameters can help to get more markers from the collected data, such as increasing the ray intersection factor, but it also can introduce more error in the reconstructed trajectories.

In one of our protocols there is a lot of marker drop out on one side of the body due to camera placement constraints, and the auto-labeling does not work well on that data. We have done our best to optimize the camera placement and adjusted the reconstruction parameters to decrease marker drop out without greatly increasing marker error. Despite all of these attempts the auto-labeling does not work on overground walking trials, but works great on standing balance tests. When using the Vicon system during treadmill walking; the camera placement is improved, marker drop out is reduced, and the standard reconstruct and label pipeline works great without modifications. Not sure if that helps, good luck.

Eric

[Arick Auyang]

Re: Vicon Nexus auto-labelling help

Problem resolved thanks to Vicon support! Thanks for responses and to Vicon for the help.

Arick

[Ton van den Bogert]

Re: Vicon Nexus auto-labelling help

Arick,

Can you please summarize how the problem was resolved?

Ton van den Bogert
Biomch-L co-moderator

[Arick Auyang]

Re: Vicon Nexus auto-labelling help

The summarize my solution/improvements:
1) This helped the most. I failed to mention that our protocol requires a few extra custom markers to be placed on the shoes. However, these markers were not originally going to be digitized. Either way those, the existance of these markers confused the system. I created a new marker file, built on plug in gait, that included our additional markers, and it worked much better. I should note that these markers not only need to be added to be labeled, but had to be attached to an existing segment or new segment.

2) Instead of using the "Autolabel Trial" option in the pipeline, I was instructed to use the "Core Processor" set to the " Label" processing level. This also improved my autolabeling luck immensely.

In fact, I am having much better luck overall. That being said, I still ahve issues where the left and right legs are flipped alot. Sometimes just the feet, sometimes the whole leg. Playing with the autolabel options can improve sometimes, but not consistently. 50% success rate still saves 50% of my time though! If anyone has suggestions on why Vicon flips the legs, I'm all ears. It is labeled correctly in the static, and even in the dynamic, all the markers are labeled correctly in terms of position. Just all the left side markers are mistaken for right and vice versa. Very strange. So close yet so far away.

Arick

Does anyone have good success using the auto-label feature with the Vicon Nexus 1.7.1 motion capturing software? I am now using the new Nexus system but came from the old Vicon Workstation software. While some things are still similar, I am not having success with Nexus. The following is what we are currently doing. We collect a static trial with the subject standing in anatomical position. We use the vicon marker set and full body plug in gait model. I label all the markers. And after this, I have been using a bit of trial and error. I try to fill gaps, run static gait, create autolabel static calibration. In the dynamic trials (running indoors across a runway), when I use the auto label function in the pipeline, it just makes a mess. I've resorted to just manually labeling, but the sheer volume of data has warranted me a few minutes to post here and request help. If you have success using the standard full body vicon marker set with the plug-in gait model with auto-labeling feature during dynamic trials, I would appreciate your insight into which steps you use in the pipeline during the static and dynamic trials and any "magic" settings you have found to work well in the auto labeling or other pipeline steps. Thanks!

Arick

[Paulo Lucareli]

Re: Vicon Nexus auto-labelling help

Dear Arick,

Try to put the thigh and shank markers assimetrical. When we place both sides simetric nexus flip a lot.

Best

Paulo Lucareli
Brazil

The summarize my solution/improvements:
1) This helped the most. I failed to mention that our protocol requires a few extra custom markers to be placed on the shoes. However, these markers were not originally going to be digitized. Either way those, the existance of these markers confused the system. I created a new marker file, built on plug in gait, that included our additional markers, and it worked much better. I should note that these markers not only need to be added to be labeled, but had to be attached to an existing segment or new segment.

2) Instead of using the "Autolabel Trial" option in the pipeline, I was instructed to use the "Core Processor" set to the " Label" processing level. This also improved my autolabeling luck immensely.

In fact, I am having much better luck overall. That being said, I still ahve issues where the left and right legs are flipped alot. Sometimes just the feet, sometimes the whole leg. Playing with the autolabel options can improve sometimes, but not consistently. 50% success rate still saves 50% of my time though! If anyone has suggestions on why Vicon flips the legs, I'm all ears. It is labeled correctly in the static, and even in the dynamic, all the markers are labeled correctly in terms of position. Just all the left side markers are mistaken for right and vice versa. Very strange. So close yet so far away.

Arick

[Ton van den Bogert]

Re: Vicon Nexus auto-labelling help

Paolo and Arick,

Thanks for your contributions, but... Please do not quote entire previous postings. I had already removed this in Arick's second posting (above) but I do not want to keep doing this. If you use "Reply With Quote", the idea is that you edit the quote, so you only quote those words that you are responding to. If you do not need to quote, use the other Reply button.

Ton van den Bogert
Biomch-L co-moderator

[Erin Bill]

Re: Vicon Nexus auto-labelling help

You could also add an additional asymmetry marker to one side ( but you would have to make sure to add it to your segment & .vst) Just place an extra marker on one side/leg and not the other, of course this marker wouldnt be used in your model for any calculations, but it would help with the flipping of labels.

[Edmund Cramp]

Re: Vicon Nexus auto-labelling help

If anyone has suggestions on why Vicon flips the legs, I'm all ears. It is labeled correctly in the static, and even in the dynamic, all the markers are labeled correctly in terms of position. Just all the left side markers are mistaken for right and vice versa. Very strange. So close yet so far away.

As Paulo mentions - an asymmetrical marker set will probably solve this problem for you - try adding just a couple of "magic" markers on one side of the body. As with your extra foot markers, the added magic asymmetrical marker(s) should not need to be labeled or even anatomically correctly placed. However they will help the software during the auto-labeling process determine which side is left and which is right on the subject.

[Arick Auyang]

Re: Vicon Nexus auto-labelling help

Thanks for the responses everyone. I use asymmetrical markers already. I have thought about adding additional markers dummy markers on one side for this purpose. I'm glad to hear encouragement to try this. I will try this on our next collection. Thanks!