3DMOTFormer: Graph Transformer for Online 3D Multi-Object Tracking

TL;DR

3DMOTFormer introduces a graph transformer-based framework for online 3D multi-object tracking that improves data association accuracy and generalizes across detectors, leveraging an online training strategy to address distribution mismatch.

Contribution

The paper presents a novel graph transformer architecture with online training for 3D MOT, enhancing data association and robustness over existing methods.

Findings

Achieves 71.2% AMOTA on nuScenes validation set.

Generalizes well across different object detectors.

Outperforms traditional non-learned algorithms in accuracy.

Abstract

Tracking 3D objects accurately and consistently is crucial for autonomous vehicles, enabling more reliable downstream tasks such as trajectory prediction and motion planning. Based on the substantial progress in object detection in recent years, the tracking-by-detection paradigm has become a popular choice due to its simplicity and efficiency. State-of-the-art 3D multi-object tracking (MOT) approaches typically rely on non-learned model-based algorithms such as Kalman Filter but require many manually tuned parameters. On the other hand, learning-based approaches face the problem of adapting the training to the online setting, leading to inevitable distribution mismatch between training and inference as well as suboptimal performance. In this work, we propose 3DMOTFormer, a learned geometry-based 3D MOT framework building upon the transformer architecture. We use an Edge-Augmented Graph Transformer to reason on the track-detection bipartite graph frame-by-frame and conduct data association via edge classification. To reduce the distribution mismatch between training and inference, we propose a novel online training strategy with an autoregressive and recurrent forward pass as well as sequential batch optimization. Using CenterPoint detections, our approach achieves 71.2% and 68.2% AMOTA on the nuScenes validation and test split, respectively. In addition, a trained 3DMOTFormer model generalizes well across different object detectors. Code is available at: https://github.com/dsx0511/3DMOTFormer.