Learning data association without data association: An EM approach to neural assignment prediction

TL;DR

This paper introduces an EM-based method for training neural data association models in multi-object tracking without requiring labeled identity data, enabling unsupervised learning of assignment matrices.

Contribution

It presents a novel EM approach using Sinkhorn networks to learn data association without supervision, improving flexibility in multi-object tracking.

Findings

Networks trained with this method can be reused in tracking tasks.

The approach does not require labeled identity data for training.

It effectively maximizes the likelihood of trajectory observations.

Abstract

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.