So, what module would be suggested to obtain distance between multiple objects, save the data. And then observe another data set of multiple objects and compare them between each other, and find missing objects in pattern and objects that don’t meet spacing of first sample? How would you go about something like this?
Observing this visually through a video output real time, and having visual indicators of missing or mis-spaced objects, for extra points.
It sounds like a detection+tracking problem.
Please check if our Deepstream SDK can meet your requirement.
Assuming the detector can catch all the objects, a tracking algorithm will be applied for the detected bbox.
The tracker will search for the identical objects in the next frame based on the given bbox.
So you will have the corresponding bounding box information between two adjacency frames.
By analyzing this information, you can find the missing ID or an object with large movement.