The Tangram Vision Platform takes a Plex at face value; we never assume to know your system better than you do. However, this can lead to some awkward situations when trying to profile your system. Here's the same RealSense™ D435i that we've been looking at with only a few constraints added:
Instead of one fully connected Plex, we have many subplexes.
Funny enough, these isolated and connected subgraphs are called components in graph theory. For obvious reasons, we choose to use the word subplex here. Sorry, graph fans.
This can be interpreted by the Tangram Vision Platform a few ways:
- We don't care about the constraints between certain components
- We do care about the constraints between all components, but didn't have enough information to form that connection
- We forgot to add the constraints that we know are present and important
In any case, this will inform the Tangram Vision Platform's behavior. Calibration, for instance, will look to fill in all explicit constraints between components. If two explicitly constrained components do not share observations (i.e. they don't see the same thing at the same time), then calibration will look to connect the components through the known constraints that have the lowest and most consistent covariance measurements.
Advanced Topic - Covariance
Covariance measurements are created by default in the calibration pipeline. It's not necessary to understand how they work to use the Tangram Vision Platform, but their values can be a rich source of information on the state of your Plex. Learn more about covariance in Advanced Topics.