CVPR Findings2026
MOSSTrack : Modality-Specific Spatio-Temporal Context Learning for RGB-T Tracking
Yisong Liu, He Yao, Junlong Cheng, Yujie Lu, Junqi Bai, Min Zhu
Abstract
Effectively leveraging spatio-temporal information is crucial for RGB-T trackers to maintain stable performance in dynamic scenarios. Most existing methods adopt identical spatio-temporal modeling strategies for both modalities and adapt to target variations through sparse imagelevel associations. However, such designs ignore modalityspecific visual cues, and image-level updates struggle to provide precise target representations. To address these issues, we propose MOSSTrack, a novel RGB-T tracker that leverages modality-specific visual cues to guide the generation of feature-level spatio-temporal information. The generated information is integrated into feature representation learning and cross-modal fusion to provide more accurate target references. Specifically, we introduce a Spatio-Temporal Information Generator (STIG), which employs learnable modality-specific tokens to select representative visual features for each modality. These features are propagated across frames to establish robust spatio-temporal associations. Furthermore, we design a simple yet effective Spatio-Temporal Context Fusion (STCF) module. It leverages spatio-temporal cues to refine target-related features and facilitates efficient cross-modal interaction. Extensive experiments on four challenging RGB-T tracking benchmarks demonstrate that MOSSTrack achieves state-of-theart performance, validating the effectiveness of our approach. Code is available at https://github.com/ lyssyl157/MoSSTrack
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- Bi-directional Adapter for Multimodal Tracking cites Learning Spatio-Temporal Transformer for Visual Tracking
- Bi-directional Adapter for Multimodal Tracking cites Visible-Thermal UAV Tracking: A Large-Scale Benchmark and New Baseline
- Learning Spatio-Temporal Transformer for Visual Tracking cites Learning the Model Update for Siamese Trackers
- Visible-Thermal UAV Tracking: A Large-Scale Benchmark and New Baseline cites Learning Spatio-Temporal Transformer for Visual Tracking
- Temporal Adaptive RGBT Tracking with Modality Prompt cites Learning the Model Update for Siamese Trackers
- Temporal Adaptive RGBT Tracking with Modality Prompt cites Learning Spatio-Temporal Transformer for Visual Tracking
- Temporal Adaptive RGBT Tracking with Modality Prompt cites Visible-Thermal UAV Tracking: A Large-Scale Benchmark and New Baseline
Builds on20
- Learning Spatio-Temporal Transformer for Visual TrackingBin Yan, Houwen Peng, Jianlong Fu, Dong Wang et al.ICCV 2021 · 1,062 citations
- Learning the Model Update for Siamese TrackersLichao Zhang, Abel Gonzalez-Garcia, Joost van de Weijer, Martin Danelljan et al.ICCV 2019 · 371 citations
- Visible-Thermal UAV Tracking: A Large-Scale Benchmark and New BaselinePengyu Zhang, Jie Zhao, Dong Wang, Huchuan Lu et al.CVPR 2022 · 225 citations
- Bi-directional Adapter for Multimodal TrackingBing Cao, Junliang Guo, Pengfei Zhu, Qinghua HuAAAI 2024 · 153 citations
- Temporal Adaptive RGBT Tracking with Modality PromptHongyu Wang, Xiaotao Liu, Yifan Li, Meng Sun et al.AAAI 2024 · 92 citations
Related papers
- Exploiting Multimodal Spatial-temporal Patterns for Video Object TrackingXiantao Hu, Ying Tai, Xu Zhao, Chen Zhao et al.AAAI 2025 · 65 citations
- Exploring Modality-Aware Fusion and Decoupled Temporal Propagation for Multi-Modal Object TrackingShilei Wang, Pujian Lai, Dong Gao, Jifeng Ning et al.AAAI 2026
- CSTrack: Enhancing RGB-X Tracking via Compact Spatiotemporal FeaturesXiaokun Feng, Dailing Zhang, Shiyu Hu, Xuchen Li et al.ICML 2025
- Prompting for Multi-Modal TrackingJinyu Yang, Zhe Li, Feng Zheng, Ales Leonardis et al.ACM MM 2022 · 167 citations
- AlignTrack: Top-Down Spatiotemporal Resolution Alignment for RGB-Event Visual TrackingChuanyu Sun, Jiqing Zhang, Yang Wang, Yuanchen Wang et al.AAAI 2026
