发表机构
National University of Science and Technology “Politehnica” Bucharest; “Simion Stoilow” Institute of Mathematics of the Romanian Academy; NORCE Norwegian Research Centre AS（布加勒斯特理工国立大学; 罗马尼亚科学院“西米翁·斯托伊洛夫”数学研究所; 挪威NORCE研究中心）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究推出Drones2BodyLanguage数据集，结合轻量几何世界模型训练12种架构，可提升空中机器人对人类交际意图的感知准确率，为社交智能无人机提供数据支撑与方法。
AI 中文摘要
远程感知人类动作与意图是具备社交智能的空中机器人的前提，但用于学习该能力的数据几乎不存在。我们推出Drones2BodyLanguage数据集，将人类动作与真实无人机镜头关联：把体现10种交际意图的化身置于未修改的4K无人机场景中，其位置、尺度和方向均符合度量标准，且在数百帧相机运动中保持稳定。实现这一目标的是本地场景的轻量几何世界模型——通过流单目深度将语义选定的锚点提升至3D空间，在该模型中，放置点被预测为具有可证刚性不变权重的仿射锚点组合，并在SVD拟合的地面旋转下重新渲染。在场景和动作不相交的划分上，针对12种架构，使用放置数据训练后，对于真实、重定向和生成的动作，平均意图准确率均大幅提升，且在两个野外场景中也验证了该增益。
英文摘要
Perceiving human motion and intent at long range is a prerequisite for socially intelligent aerial robots, yet the data to learn it barely exists. We introduce Drones2BodyLanguage, a dataset grounding human motion in real UAV footage: avatars manifesting ten communicative intents are placed into unmodified 4K drone scenes with metrically correct position, scale and orientation, maintained over hundreds of frames of camera motion. Enabling it is a lightweight geometric world model of the local scene - semantically selected anchors lifted to 3D through streaming monocular depth - in which a placement point is predicted as an affine anchor combination with provably rigid-invariant weights, and re-rendered under an SVD-fitted ground rotation. Across twelve architectures on scene- and motion-disjoint splits, training on placed data lifts mean intent accuracy by a wide margin for real, retargeted and generated motion alike, with gains confirmed on two in-the-wild scenes.
