Dual-level Adaptation for Multi-Object Tracking: Building Test-Time Calibration from Experience and Intuition
Wen Guo, Pengfei Zhao, Zongmeng Wang, Yufan Hu, Junyu Gao · 2026
Read Dual-level Adaptation for Multi-Object Tracking: Building Test-Time Calibration from Experience and Intuition by Wen Guo, Pengfei Zhao, Zongmeng Wang, Yufan Hu, Junyu Gao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-21629.txt.
Description
Guo and co-authors adapt multi-object tracking to new test conditions while the tracker runs. Appearance and motion shifts between training and test data usually drag performance down. Frame-level adaptation alone misses identity links across time. Their TCEI framework pairs two systems modeled on human judgment. An intuitive system recalls recent objects for fast predictions. An experiential system uses lessons from earlier test videos to calibrate those calls. Confident and uncertain objects become priors and reflective cases for later frames. Benchmark tests show steady gains under distribution shift.
Catalog record
| CREATOR | Wen Guo, Pengfei Zhao, Zongmeng Wang, Yufan Hu, Junyu Gao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37696 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.21629