AI 中文总结
该研究针对目标状态定位问题提出无需训练的 Déjà Cue 框架，通过词汇相对坐标系校准图像-文本相似度，在78个 VOST 序列上显著提升了 R@1 和 Top-1 tIoU 指标。
AI 中文摘要
跟踪（Tracking）通过视觉变化关联同一目标的观测结果，但本身无法确定目标何时为空或被填充、完整或被切割。我们提出身份条件化的状态-时刻检索任务：给定一个已跟踪目标的历史序列和多个备选状态描述，定位每个描述的状态成立的时间区间。绝对图像-文本相似度会独立评估各个描述；由于每个可见帧都描绘同一目标，共享的目标兼容性会掩盖识别目标区间所需的状态证据。这些备选描述提供了缺失的参考：应针对其他状态衡量某一状态的证据。我们提出 Déjà Cue，这是一个无需训练的框架，可将这些备选描述转化为词汇相对坐标系。该框架从每个描述中减去其状态平衡质心，校准帧得分，并使用冻结编码器在连续可见序列内扫描多个持续时间。在78个 VOST 历史序列上，固定时间扫描设置仅改变查询参考，R@1在 tIoU 0.5 时从10.3%提升至20.5%，几乎翻倍；Top-1 tIoU 从16.0%提升至21.5%。候选排名分析显示，词汇相对查询在同一候选集中将有用区间排名更高，因此相关状态描述可作为目标特定的查询时坐标系，用于读取冻结的视觉表征。
英文摘要
Tracking links observations of the same object through visual change, yet cannot by itself determine when the object is empty or filled, intact or cut. We formulate identity-conditioned state-moment retrieval: given a tracked-object history and alternative state descriptions, localize an interval in which each described state holds. Absolute image-text similarity scores descriptions independently; because every visible frame depicts the same target, shared object compatibility can obscure the state evidence needed to identify the target interval. The alternatives provide the missing reference: evidence for one state should be measured against the others. We introduce Déjà Cue, a training-free framework that turns these alternatives into a vocabulary-relative coordinate system. It subtracts their state-balanced centroid from each description, calibrates frame scores, and scans multiple durations within contiguous visible runs using a frozen encoder. On 78 VOST histories, holding the temporal scan fixed and changing only the query reference nearly doubles R@1 at tIoU 0.5 from 10.3\% to 20.5\% and raises Top-1 tIoU from 16.0\% to 21.5\%. Candidate-rank analyses show that vocabulary-relative queries rank useful intervals higher within the same candidate set. Related state descriptions can therefore serve as an object-specific, query-time coordinate system for reading frozen visual representations.
CommentsCode available at https://github.com/HaofanCao/DejaCue
