研究人员开发了 H-Flow，一种用于从单目视频估计人体场景流的新型自监督方法。该方法整合了骨骼运动学和表面变形，克服了现有模型的局限性。H-Flow 利用受物理启发的先验和一个新的合成基准 DynAct4D，实现了最先进的性能，并能泛化到真实世界的视频。 AI
影响 引入了一种新的密集人体运动估计方法，有望改进动画和虚拟现实应用。
排序理由 该集群包含一篇详细介绍新模型和基准的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
H-Flow 模型利用物理先验从视频中估计人体运动
报道来源 [2]
-
H-Flow: Self-supervised Human Scene Flow via Physics-inspired Joint Multi-modal Learning
arXiv:2605.22629v1 Announce Type: new Abstract: Parametric human models capture global pose but cannot represent the non-rigid surface dynamics of clothing and soft tissue. Generic scene flow estimates dense motion but breaks down on articulated bodies, where pixel-level supervis…
-
H-Flow: Self-supervised Human Scene Flow via Physics-inspired Joint Multi-modal Learning
Parametric human models capture global pose but cannot represent the non-rigid surface dynamics of clothing and soft tissue. Generic scene flow estimates dense motion but breaks down on articulated bodies, where pixel-level supervision is also intractable to acquire. We introduce…