面向多目标跟踪的双层适应：从经验与直觉构建测试时校准
Dual-level Adaptation for Multi-Object Tracking: Building Test-Time Calibration from Experience and Intuition
把 TTA 从逐帧适应扩展到跨帧、跨视频的时序与身份关联层面：用瞬时记忆做快速关联预测，再用历史测试视频积累的经验对预测做再校准，可直接迁移到任何在线跟踪器（如 ByteTrack、BoT-SORT）的推理阶段以缓解外观/运动/类别分布漂移。
原文依据 they primarily focus solely on frame-level adaptation while neglecting temporal consistency and identity association across frames and videos
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论多目标跟踪（MOT）长期以来是计算机视觉中的基础任务，在各类真实场景中有着广泛应用。然而，由于训练数据与测试数据之间在外观、运动模式和类别上存在分布偏移，模型在 MOT 的在线推理过程中性能会显著下降。测试时自适应（TTA）已成为缓解此类分布偏移的一种有前景的范式。但现有 TTA 方法在 MOT 中往往难以取得令人满意的结果，因为它们主要只关注帧级适应，而忽略了跨帧、跨视频的时序一致性与身份关联。受人类决策过程的启发，本文提出了一种从经验与直觉进行测试时校准（TCEI）的框架。在该框架中，直觉系统利用瞬时记忆召回近期观察到的目标以进行快速预测，而经验系统则利用从先前测试视频中积累的经验对这些直觉预测进行重新评估与校准。此外，在线测试过程中置信与不确定的目标分别被用作历史先验和反思案例，使模型能够适应测试环境并缓解性能退化。大量实验表明，所提出的 TCEI 框架在多个基准数据集上持续取得更优性能，并显著增强了模型在分布偏移下的适应能力。代码将发布于 https://github.com/1941Zpf/TCEI。
英文原文摘要 1586 字符
Multiple Object Tracking (MOT) has long been a fundamental task in computer vision, with broad applications in various real-world scenarios. However, due to distribution shifts in appearance, motion pattern, and catagory between the training and testing data, model performance degrades considerably during online inference in MOT. Test-Time Adaptation (TTA) has emerged as a promising paradigm to alleviate such distribution shifts. However, existing TTA methods often fail to deliver satisfactory results in MOT, as they primarily focus solely on frame-level adaptation while neglecting temporal consistency and identity association across frames and videos. Inspired by human decisionmaking process, this paper propose a Test-time Calibration from Experience and Intuition (TCEI) framework. In this framework, the Intuitive system utilizes transient memory to recall recently observed objects for rapid predictions, while the Experiential system leverages the accumulated experience from prior test videos to reassess and calibrate these intuitive predictions. Furthermore, both confident and uncertain objects during online testing are exploited as historical priors and reflective cases, respectively, enabling the model to adapt to the testing environment and alleviate performance degradation. Extensive experiments demonstrate that the proposed TCEI framework consistently achieves superior performance across multiple benchmark datasets and significantly enhances the model's adaptability under distribution shifts. The code will be released at https://github.com/1941Zpf/TCEI.