SiamMOT: Siamese Multi-Object Tracking

TL;DR

SiamMOT is a region-based Siamese network for online multi-object tracking that models object motion explicitly or implicitly, significantly improving tracking accuracy and efficiency across multiple datasets.

Contribution

The paper introduces SiamMOT, a novel Siamese multi-object tracking network with explicit and implicit motion models, outperforming state-of-the-art methods and running efficiently in real-time.

Findings

Outperforms state-of-the-art on MOT17, TAO-person, and Caltech datasets.

Outperforms ACM MM'20 HiEve Grand Challenge winners.

Runs at 17 FPS on 720P videos on a single GPU.

Abstract

In this paper, we focus on improving online multi-object tracking (MOT). In particular, we introduce a region-based Siamese Multi-Object Tracking network, which we name SiamMOT. SiamMOT includes a motion model that estimates the instance's movement between two frames such that detected instances are associated. To explore how the motion modelling affects its tracking capability, we present two variants of Siamese tracker, one that implicitly models motion and one that models it explicitly. We carry out extensive quantitative experiments on three different MOT datasets: MOT17, TAO-person and Caltech Roadside Pedestrians, showing the importance of motion modelling for MOT and the ability of SiamMOT to substantially outperform the state-of-the-art. Finally, SiamMOT also outperforms the winners of ACM MM'20 HiEve Grand Challenge on HiEve dataset. Moreover, SiamMOT is efficient, and it runs at 17 FPS for 720P videos on a single modern GPU. Codes are available in \url{https://github.com/amazon-research/siam-mot}.