Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.02654 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917382605242368 |
|---|---|
| author | Huang, Yuqing Lin, Liting Zhuang, Weijun He, Zhenyu Li, Xin |
| author_facet | Huang, Yuqing Lin, Liting Zhuang, Weijun He, Zhenyu Li, Xin |
| contents | Temporal information is crucial for visual tracking, but existing multi-frame trackers are vulnerable to model drift caused by naively aggregating noisy historical predictions. In this paper, we introduce DTPTrack, a lightweight and generalizable module designed to be seamlessly integrated into existing trackers to suppress drift. Our framework consists of two core components: (1) a Temporal Reliability Calibrator (TRC) mechanism that learns to assign a per-frame reliability score to historical states, filtering out noise while anchoring on the ground-truth template; and (2) a Temporal Guidance Synthesizer (TGS) module that synthesizes this calibrated history into a compact set of dynamic temporal priors to provide predictive guidance. To demonstrate its versatility, we integrate DTPTrack into three diverse tracking architectures--OSTrack, ODTrack, and LoRAT-and show consistent, significant performance gains across all baselines. Our best-performing model, built upon an extended LoRATv2 backbone, sets a new state-of-the-art on several benchmarks, achieving a 77.5% Success rate on LaSOT and an 80.3% AO on GOT-10k. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_02654 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Drift-Resilient Temporal Priors for Visual Tracking Huang, Yuqing Lin, Liting Zhuang, Weijun He, Zhenyu Li, Xin Computer Vision and Pattern Recognition Temporal information is crucial for visual tracking, but existing multi-frame trackers are vulnerable to model drift caused by naively aggregating noisy historical predictions. In this paper, we introduce DTPTrack, a lightweight and generalizable module designed to be seamlessly integrated into existing trackers to suppress drift. Our framework consists of two core components: (1) a Temporal Reliability Calibrator (TRC) mechanism that learns to assign a per-frame reliability score to historical states, filtering out noise while anchoring on the ground-truth template; and (2) a Temporal Guidance Synthesizer (TGS) module that synthesizes this calibrated history into a compact set of dynamic temporal priors to provide predictive guidance. To demonstrate its versatility, we integrate DTPTrack into three diverse tracking architectures--OSTrack, ODTrack, and LoRAT-and show consistent, significant performance gains across all baselines. Our best-performing model, built upon an extended LoRATv2 backbone, sets a new state-of-the-art on several benchmarks, achieving a 77.5% Success rate on LaSOT and an 80.3% AO on GOT-10k. |
| title | Drift-Resilient Temporal Priors for Visual Tracking |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.02654 |