生成式点跟踪与预测
Generative Point Tracking and Forecasting
把点跟踪与运动预测统一为视频条件下的因果流匹配生成问题，通过切换视觉信号在跟踪与预测间切换，为长时点轨迹建模提供了可复用的生成式框架。
原文依据 We train a causal, video-conditioned flow matching model to predict point positions.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论运动预测预测点在未来将移动到何处，而运动跟踪预测点当前位于何处。尽管二者存在这些相似之处，现有方法对这两个问题的处理却相当不同。本文提出一个能够同时解决这两项任务的统一模型。我们训练一个因果的、以视频为条件的流匹配模型来预测点位置。所得模型只需改变其视觉信号，即可轻松在点跟踪与预测之间切换。尽管模型简单，我们发现它在点预测上优于先前工作，并在 TAP-Vid 基准上取得与最优方法相当的性能。
英文原文摘要 667 字符
Motion forecasting predicts where points will move in the future, while motion tracking predicts where they are in the present. Despite these similarities, existing approaches to the two problems are quite different. In this paper, we propose a unified model that can address both tasks. We train a causal, video-conditioned flow matching model to predict point positions. The resulting model can easily toggle between point tracking to forecasting by changing its visual signal. Despite our model's simplicity, we find that it outperforms prior work in point forecasting and obtains performance that is competitive with the state-of-the-art on the TAP-Vid benchmark.