Summary: Patch 7 of Theia 2023 is now live with some modest but helpful features! These features are located in the settings pane and provide added clarity on how people are identified and tracked within the volume.
Okay, let’s dive in!
1. Bounding Box: We've added a bounding box that restricts the 3D volume of analysis. This is really helpful if you have a person in the background that is being tracked, but is not part of the study. Without this bounding box, this person will add additional skeletons and c3d files to the analysis that are typically not desired and require manual assessment to determine which file is the actual participant. You can also use the camera locations to define this box (or a starting point). So, if in that frame, part of the body is in the box, it will track it, otherwise it will be removed.
2. Remove stale IDs: This option removes person tracking when a person has not been tracked for at least one hundred frames. This is predominantly useful if the person is entering and leaving the volume, and you would like each session in the volume to be a single c3d file, instead of a longer, but more sparse, single c3d file. So, being extremely explicit about this - if I am walking across the lab and entering, then exiting the camera views, if this option is on, the analysis will yield N walk c3d files instead of one.
3. Person tracking mode: Person tracking defines how the order of person IDs are determined (So, which person is person 0, 1..., N). We have added three options to provide a bit more flexibility in analysis: most visible, closest to origin, and order of appearance:
Most visible: After every person is identified, we divide them into two groups - those that are seen in more than 75% of the total frames/views and those that are not. Then we sort these two groups with respect to distance from the origin.
Closest to the origin: Sorted by the minimum distance that that individual was from the origin of the volume throughout the entire trial regardless of visibility. So, if a person walks directly over the origin, they will likely be person 0.
Order of appearance: Identified by the order which they appear in the volume.
In practice, these tools can be applied to a variety of circumstances, however there is one example that drove a lot of these features. In this particular collection, the goal was to collect an entire team and time was of the essence. So, one by one, each player ran into the volume, performed a maneuver, and left the volume. The cameras were recording the entire time, to avoid missing data and slowing down the collection. The desired output from the analysis was a single c3d file, for each person, in the order that they entered, while removing any superfluous people in the background (see, this is where the settings come in!). To achieve this, the experimenters would need to click Remove stale IDs, set the person tracking mode to order of appearance, and define a tracking bounding box to remove the observers.
I admit, this may not be my most exciting and inspiring blog for you, but for us, this is really what it’s all about. Small features, driven by the requirements of particular studies, that ultimately save time and effort for the experimenter. We want users spending their time defining experiments, interpreting data, writing articles, sharing this with the world, and helping individuals move better, and not labor over getting usable data. If this type of thinking is in line with yours, send us an email and we can provide you with more information.
To learn more about Theia, click here to book a demo.
Commenti