Multi-Object Tracking in Production Environments

Multi-object tracking has many uses cases in autonomous driving, robotics and security. Tracking is important as it allows us reason about the dynamic world and make actionable decisions based-off predicted object trajectories. As an example, in autonomous driving, not only the location of objects is important but also their predicted future trajectories are needed in order to make informed decisions. It is important to understand the hardware constraints in order to develop trackers that work in real-time. Understanding the necessary camera frame rate needed to perform actionable real-time tracking is an example of such hardware constraint. For robotics, developing adequate tracking solutions that satisfy hardware constraint will allow the verification that objects are moving as they should.

Faculty Supervisor:

Florian Shkurti

Student:

Partner:

Kindred AI

Discipline:

Computer science

Sector:

Artificial Intelligence

University:

University of Toronto

Program: