How To Train Your Deep Multi-Object Tracker

TL;DR

This paper introduces a differentiable framework for training deep multi-object trackers end-to-end by approximating tracking evaluation metrics, leading to improved performance and new state-of-the-art results on MOTChallenge.

Contribution

It proposes a novel differentiable proxy for MOTA and MOTP, enabling direct optimization of tracking performance in deep learning models.

Findings

Achieves state-of-the-art results on MOTChallenge

Improves existing multi-object tracking methods

Provides a new differentiable training framework

Abstract

The recent trend in vision-based multi-object tracking (MOT) is heading towards leveraging the representational power of deep learning to jointly learn to detect and track objects. However, existing methods train only certain sub-modules using loss functions that often do not correlate with established tracking evaluation measures such as Multi-Object Tracking Accuracy (MOTA) and Precision (MOTP). As these measures are not differentiable, the choice of appropriate loss functions for end-to-end training of multi-object tracking methods is still an open research problem. In this paper, we bridge this gap by proposing a differentiable proxy of MOTA and MOTP, which we combine in a loss function suitable for end-to-end training of deep multi-object trackers. As a key ingredient, we propose a Deep Hungarian Net (DHN) module that approximates the Hungarian matching algorithm. DHN allows estimating the correspondence between object tracks and ground truth objects to compute differentiable proxies of MOTA and MOTP, which are in turn used to optimize deep trackers directly. We experimentally demonstrate that the proposed differentiable framework improves the performance of existing multi-object trackers, and we establish a new state of the art on the MOTChallenge benchmark. Our code is publicly available from https://github.com/yihongXU/deepMOT.