研究人员探讨了时间分辨率如何影响人类动作的零样本语义理解能力，特别是针对快速运动。他们以剑道为案例进行研究，发现更高的帧率显著提高了预训练视频语言模型在无特定任务训练的情况下语义区分动作的能力。研究结果表明，高速感知增强了动作识别的可解释性和稳定性，这对于人机交互等应用至关重要。 AI
影响 增强了对快速、细微动作的零样本动作识别能力，改善了人机交互。
排序理由 学术论文，详细介绍了一种新颖的零样本动作识别方法。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2605.00496v1 Announce Type: new Abstract: Understanding human actions from visual observations is essential for human--robot interaction, particularly when semantic interpretation of unfamiliar or hard-to-annotate actions is required. In scenarios such as rapid and less com…
Understanding human actions from visual observations is essential for human--robot interaction, particularly when semantic interpretation of unfamiliar or hard-to-annotate actions is required. In scenarios such as rapid and less common activities, collecting sufficient labeled da…