Learning data association without data association: An EM approach to neural assignment prediction
This addresses the challenge of data association for multi-object tracking systems by eliminating the need for costly labeled data, though it is incremental as it builds on existing neural and assignment methods.
The paper tackles the problem of training neural models for data association in multi-object tracking without requiring labeled data, by introducing an expectation maximization approach that trains a Sinkhorn network to predict assignment matrices, achieving results that enable re-use in downstream tracking applications.
Data association is a fundamental component of effective multi-object tracking. Current approaches to data-association tend to frame this as an assignment problem relying on gating and distance-based cost matrices, or offset the challenge of data association to a problem of tracking by detection. The latter is typically formulated as a supervised learning problem, and requires labelling information about tracked object identities to train a model for object recognition. This paper introduces an expectation maximisation approach to train neural models for data association, which does not require labelling information. Here, a Sinkhorn network is trained to predict assignment matrices that maximise the marginal likelihood of trajectory observations. Importantly, networks trained using the proposed approach can be re-used in downstream tracking applications.