Skip to main content
Models & Technology

Noitom Robotics Releases HiPHI, Open-Sources 617.5 Hours of High-Precision Human Motion Data

Noitom Robotics has released the HiPHI dataset, totaling 617.5 hours and containing high-precision full-body motion and human-object interaction data. The dataset is open-sourced on Hugging Face, and models trained on it have enabled actions such as running and crawling on the Unitree G1 robot. #OpenSourceRobotics#

IT HomeAbout 3 min read
Noitom Robotics releases HiPHI with 617.5 hours of high-precision human motion data

Noitom Robotics released HiPHI during the 2026 World Robot Conference. It is a high-precision optical motion-capture dataset designed for researchers and engineers working in humanoid robot learning, digital humans, and computer graphics.

Noitom Robotics releases HiPHI with 617.5 hours of high-precision human motion data

According to reports, the dataset totals 617.5 hours, including 371.8 hours of full-body human motion data and 245.7 hours of human-object interaction data, with left-right mirroring. The data was collected from 132 motion-capture performers at a frequency of 90 Hz with submillimeter optical motion-capture precision. It was structurally organized into motion units based on FrameNet's framework for organizing the semantics of human actions.

Noitom Robotics releases HiPHI with 617.5 hours of high-precision human motion data

The dataset is now publicly available on Hugging Face, with documentation, a paper, and other resources. Models trained on HiPHI have been deployed on a physical Unitree G1 humanoid robot, enabling actions such as running, sitting down, crawling, carrying boxes, and pulling suitcases.

Reference:

noitomrobotics/HiPHI · Datasets at Hugging Face