UBATrack: Spatio-Temporal State Space Model for General Multi-Modal Tracking

TL;DR

UBATrack introduces a novel spatio-temporal state space model for multi-modal tracking, effectively capturing cross-modal dependencies and improving robustness without extensive fine-tuning, outperforming existing methods across various benchmarks.

Contribution

The paper proposes UBATrack, a new multi-modal tracking framework utilizing a Mamba-style state space model with modules for spatio-temporal modeling and feature mixing, enhancing efficiency and performance.

Findings

Outperforms state-of-the-art on multiple benchmarks

Effectively models cross-modal dependencies

Improves training efficiency without full fine-tuning

Abstract

Multi-modal object tracking has attracted considerable attention by integrating multiple complementary inputs (e.g., thermal, depth, and event data) to achieve outstanding performance. Although current general-purpose multi-modal trackers primarily unify various modal tracking tasks (i.e., RGB-Thermal infrared, RGB-Depth or RGB-Event tracking) through prompt learning, they still overlook the effective capture of spatio-temporal cues. In this work, we introduce a novel multi-modal tracking framework based on a mamba-style state space model, termed UBATrack. Our UBATrack comprises two simple yet effective modules: a Spatio-temporal Mamba Adapter (STMA) and a Dynamic Multi-modal Feature Mixer. The former leverages Mamba's long-sequence modeling capability to jointly model cross-modal dependencies and spatio-temporal visual cues in an adapter-tuning manner. The latter further enhances multi-modal representation capacity across multiple feature dimensions to improve tracking robustness. In this way, UBATrack eliminates the need for costly full-parameter fine-tuning, thereby improving the training efficiency of multi-modal tracking algorithms. Experiments show that UBATrack outperforms state-of-the-art methods on RGB-T, RGB-D, and RGB-E tracking benchmarks, achieving outstanding results on the LasHeR, RGBT234, RGBT210, DepthTrack, VOT-RGBD22, and VisEvent datasets.