Simple Cues Lead to a Strong Multi-Object Tracker

TL;DR

This paper demonstrates that simple tracking-by-detection methods, when combined with basic appearance features and motion models, can achieve state-of-the-art multi-object tracking performance across multiple datasets.

Contribution

The authors show that a standard re-identification network combined with simple motion cues can match complex end-to-end models in multi-object tracking.

Findings

Achieves state-of-the-art results on MOT17, MOT20, BDD100k, and DanceTrack datasets.

Simple cues combined with a re-identification network are highly effective.

Analysis of failure cases provides insights for further improvements.

Abstract

For a long time, the most common paradigm in Multi-Object Tracking was tracking-by-detection (TbD), where objects are first detected and then associated over video frames. For association, most models resourced to motion and appearance cues, e.g., re-identification networks. Recent approaches based on attention propose to learn the cues in a data-driven manner, showing impressive results. In this paper, we ask ourselves whether simple good old TbD methods are also capable of achieving the performance of end-to-end models. To this end, we propose two key ingredients that allow a standard re-identification network to excel at appearance-based tracking. We extensively analyse its failure cases, and show that a combination of our appearance features with a simple motion model leads to strong tracking results. Our tracker generalizes to four public datasets, namely MOT17, MOT20, BDD100k, and DanceTrack, achieving state-of-the-art performance. https://github.com/dvl-tum/GHOST.