Overview

RoSHI is a low-cost, portable system for in-the-wild human motion capture. It fuses information from 9 body-worn IMUs and Project Aria glasses to output synchronized 3D body pose (SMPL format), egocentric RGB video, and a globally consistent root trajectory. The figure above shows an example of data being collected, processed, and deployed to a real robot.

Key Features

• Low-cost: under $500 in tracker hardware — see the Hardware section for a detailed item list

• Quick calibration: a 20–40 second iPhone video is enough — refer to the Calibration section for details

• Portable & In-the-wild: the entire system is easily wearable, and once calibrated you can move freely in any environment