Multi-Camera Multi-Vehicle Tracking With Domain Generalization and Contextual Constraints
In this paper, we develop and propose a system for Multi-Camera Multi-Target (MCMT) Vehicle Tracking in Track 1 of AI City Challenge 2022. There are many technical difficulties to the MCMT problem such as a common lack of labelled data in real scenarios, a distortion of vehicle detailed appearances in recording, and ambiguity between highly similar vehicles. Taking those into account, we develop a 3-component MCMT system that exploits vehicle behavior, leverages synthetic data and augmentation techniques to exploit as much labeled data as possible, and enforce contextual constraints to address ambiguity in terms of vehicle appearances. Specifically, our system involves a motion-driven vehicle tracker, applying MixStyle domain generalization on the TransReID model, and experiment with various constraints such as neighbour matching.