About

About MEgo

MEgo Engine turns synchronised recordings of human activity into training datasets through reconstruction, multimodal annotation, quality assessment and iterative dataset management. The capture hardware is offered as Ego View, Ego + Wrist and UMI Gripper configurations so teams can record egocentric behaviour, wrist motion or gripper-aligned manipulation trajectories without deploying a complete robot.

Across those configurations, the system records RGB, depth, inertial, tactile, audio and motion data in homes, offices, factories, warehouses and other working environments. Maniformer reported on 31 August 2026 that it had delivered 20,000 MEgo-series devices and collected more than one million hours spanning over 10,000 environments, 50,000 object categories and 500 tasks.

Operations

Operating Use
more than 10,000 environments - 2026-08-31 - real-world Physical AI data collection - Maniformer reported MEgo data collection spanning more than 10,000 real-world environments.
Source
Operating Milestone
more than 1 million hours - 2026-08-31 - real-world Physical AI data collection - Maniformer reported more than one million hours of real-world data generated through MEgo.
Source
Shipments
20,000 MEgo-series devices delivered - 2026-08-31 - real-world Physical AI data collection - Maniformer reported 20,000 MEgo-series data-collection devices delivered by Aug. 31, 2026.
Source

Timeline

Maniformer launches Mifeng Pai robot-data crowdsourcing platform
2026-09-23 - Product Launch - Participants accept data-collection tasks through the app and use MEgo wearable hardware in retail, warehouse, manufacturing and home settings. The company reports 20,000 registered users and 13,000 submissions during a one-month beta. The platform covers 22 scenario categories, more than 5,000 tasks and over 50,000 real-world environments, according to the company.
Source
Maniformer delivers 20,000th MEgo after passing one million data hours
2026-08-31 - Production Milestone - The milestone unit was delivered to JD.com for deployment in real operating environments. The data estate spans more than one million hours across over 10,000 environments, 50,000 object categories and 500 tasks. A new data partnership with Tencent Robotics X supports embodied-model development.
Source