Know Your Surroundings: Panoramic Multi-Object Tracking by Multimodality Collaboration

TL;DR

This paper introduces MMPAT, a multimodal panoramic multi-object tracking framework that combines 2D images and 3D point clouds to improve tracking accuracy in complex driving and navigation scenarios.

Contribution

It proposes a novel multimodal approach for panoramic multi-object tracking that outperforms existing methods significantly on the JRDB dataset.

Findings

Achieves top performance in detection and tracking on JRDB dataset.

Outperforms state-of-the-art methods by 15.7 AP and 8.5 MOTA.

Demonstrates robustness in complex environments with background clutter and poor lighting.

Abstract

In this paper, we focus on the multi-object tracking (MOT) problem of automatic driving and robot navigation. Most existing MOT methods track multiple objects using a singular RGB camera, which are prone to camera field-of-view and suffer tracking failures in complex scenarios due to background clutters and poor light conditions. To meet these challenges, we propose a MultiModality PAnoramic multi-object Tracking framework (MMPAT), which takes both 2D panorama images and 3D point clouds as input and then infers target trajectories using the multimodality data. The proposed method contains four major modules, a panorama image detection module, a multimodality data fusion module, a data association module and a trajectory inference model. We evaluate the proposed method on the JRDB dataset, where the MMPAT achieves the top performance in both the detection and tracking tasks and significantly outperforms state-of-the-art methods by a large margin (15.7 and 8.5 improvement in terms of AP and MOTA, respectively).