HumanTracker：面向全面且与人类对齐的运动跟踪基准
HumanTracker: Towards Comprehensive and Human-Aligned Motion Tracking Benchmark
浏览论文内容
中文总结 AI 辅助
针对人形运动跟踪评估与人类感知不符、基准多样性不足的问题，提出HumanTracker基准及偏好对齐指标HumanScore，该指标可更好预测人类偏好并揭示运动学指标遗漏的接触与稳定性故障。
中文摘要 AI 辅助
人形机器人运动跟踪是远程操作和全身模仿的核心，但评估结果常与人们在视频中的感知不符。运动学误差仅平均逐帧位姿差异，却忽略了最关键的物理缺陷，尤其是不稳定的支撑状态和错误接触，如足部滑动、触地时机不当。同时，广泛使用的测试套件规模较小，缺乏应对接触丰富、长时序行为所需的多样性。我们推出HumanTracker，使人形跟踪评估既与人类感知对齐又具备可扩展性。HumanTracker基准包含约153小时来自多名专业表演者的光学运动轨迹，分为四个运动家族，并带有文本标签用于细粒度诊断。我们进一步提出HumanScore，这是一种偏好对齐的指标，基于包含24个动作的12000个运动对训练而成。在代表性的最先进跟踪器上，HumanScore能更好地预测人类偏好，并揭示运动学指标常遗漏的接触与稳定性故障。
英文摘要
Humanoid motion tracking is central to teleoperation and whole-body imitation, yet evaluation often disagrees with what people perceive in videos. Kinematic errors average per-frame pose differences but miss the physical artifacts that matter most, particularly unstable support and incorrect contacts such as foot skating and mistimed touch-downs. Meanwhile, widely used test suites are small and lack the diversity needed to stress contact-rich, long-horizon behaviors. We introduce HumanTracker to make humanoid tracking evaluation both perceptually aligned and scalable. The HumanTracker benchmark contains approximately 153 hours of optical motion trajectories from multiple professional performers, organized into four motion families with text labels for fine-grained diagnosis. We further propose HumanScore, a preference-aligned metric trained on 12K motion pairs containing 24K motions. Across representative state-of-the-art trackers, HumanScore better predicts human preferences and reveals contact and stability failures that kinematic metrics often miss.
发表机构
- Nankai University（南开大学）
- Tsinghua University（清华大学）
- Galbot
- Shanghai Jiao Tong University（上海交通大学）
- Peking University（北京大学）
- Shanghai Qi Zhi Institute（上海启智研究院）
机构由 AI 辅助整理，请以论文原文为准。
