[Feature-Request] Robust multi-device tracking

[Erol444]

Start with the why:

Many customers have already shown interest in robust multi-device object (people, vehicle..) tracking.

Move to the what:

Develop a demo that showcases how customers can leverage a few approaches to achieve robust multi-device object tracking:

• Use OAK's ObjectTracker node
• Extract features from the object and running DeepSORT (or similar) to track the object, demo here
• Test object (people, vehicle)-specific re-identification models (that provide features), eg. people-reidentification demo, or vehicle-reid-0001 from OMZ.
• If multiple devices can see the same scene, use multi-device calibration and try to match a person across multiple devices, eg. by features, or by 3D coordiante matching.
• Share features across cameras (via host computer) so an object (eg person) is tracked through the space (eg building) even if cameras don't see the same scene

Demo could showcase a basic birdseye view floor plan, how each camera is positioned (and its FOV), and where people are/how they are moving around. Going with robothub app would likely be the best idea, and use a web technology like d3js to visualize everything.