Stories / 14 Oct 2026

A motion capture dataset called HiPHI runs to 617.5 hours, and its white paper reports policies trained on it deployed on a physical Unitree G1 humanoid.

Slide 1 of 8. 617.5 hours of motion capture, tested on a G1 The white paper for the HiPHI dataset reports policies trained on it deployed on a real Unitree G1 humanoid.

Slide 2 of 8. A whole-body motion dataset, deployed on a G1 HiPHI's white paper reports policies trained on the data deployed on a physical Unitree G1 humanoid.

Slide 3 of 8. Sub-millimeter optical capture The dataset runs to 617.5 hours, captured with optical motion capture at sub-millimeter accuracy.

Slide 4 of 8. 245.7 hours involve objects That share is human-object interaction, with synchronized object trajectories and meshes. The white paper says that pairing is what makes the data useful for teaching carrying, pushing and pulling.

Slide 5 of 8. FrameNet for coverage The paper uses FrameNet, a linguistic framework for human action, to organize coverage. It says FrameNet can guide motion capture collection to cover a broad range of whole-body motion.

Slide 6 of 8. Sponsored by Noitom Robotics The white paper is sponsored by Noitom Robotics and published with IEEE Spectrum and Wiley.

Slide 7 of 8. A year from now, you'll be glad you followed today. AI + robotics daily, plain English, no hype, always free. Every story checked against its source. Follow @dailytechwizard.

Slide 8 of 8. Screenshot of the source on content.knowledgehub.wiley.com: HiPHI: A Large-Scale Benchmark for High-Precision Human Motion and Object Interaction

A motion capture dataset called HiPHI runs to 617.5 hours, and its white paper reports policies trained on it deployed on a physical Unitree G1 humanoid. 245.7 of those hours are human-object interaction, with synchronized object trajectories and meshes. The paper also introduces a benchmark suite for measuring motion diversity and interaction grounding. Worth knowing: the white paper is sponsored by Noitom Robotics. Source: link in the first comment

Read the source: HiPHI: A Large-Scale Benchmark for High-Precision Human Motion and Object Interaction (content.knowledgehub.wiley.com)