Shanghai-based company Maniformer, which provides training data for physical AI, has introduced a crowdsourcing platform. This platform allows regular individuals to earn income by documenting everyday activities intended for robot learning. The launch took place in Shanghai on September 23, under the motto that everyone can act as a teacher, according to Leiphone.
Maniformer describes this as the first all-category crowdsourcing platform dedicated to generating high-quality data for physical AI. The CEO of Maniformer, Yao Maoqing, stated that the primary barrier preventing robots from being widely used in homes, shops, and factories is not related to algorithms or hardware, but rather the scarcity of extensive datasets featuring real human movements.
The platform is designed to convert common routines—such as arranging items, preparing tea, or folding laundry—into valuable training material. The operational structure involves two distinct levels. The foundational layer consists of infrastructure elements, including MEgo capture devices, a mobile application, and the MEgo Engine for subsequent data processing.
Specific hardware includes MEgo View, a head-mounted apparatus whose cameras capture views exceeding 300 degrees, and MEgo Gripper, which reconstructs movement with millimeter precision, offering an option for 3D touch sensing.
Above this infrastructure operates a network comprising actual workplaces and data collectors. Users must register using their true names and receive a device before they can accept a task, record it, pass a quality assessment, and receive payment. The platform ensures that withdrawals are processed within two minutes, and tax obligations are managed internally.
The platform currently features 22 different categories, encompassing over 5,000 specific task types and more than 50,000 documented real-life scenarios. During a trial period conducted between August 15 and September 15, close to 20,000 participants registered interest.
Furthermore, Maniformer initiated a robot trainer program, granting the designation to 16 beta users. The company's business vice president, Zhang Zhifu, detailed a subsidy package totaling RMB 100 million. This funding is allocated as follows: RMB 50 million for tasks, RMB 20 million for devices, RMB 30 million to establish an offline service network across 40 cities, and RMB 2 million for insurance covering both personnel and equipment.
The company has established a five-tier progression system, ranging from M1 to M5, with ambitious goals of achieving one million daily active users and one million robot trainers within a two-year timeframe. In addition to its internal operations, Maniformer is collaborating with the China Academy of Information and Communications Technology on a study concerning crowdsourced embodied AI data. The firm has also formed a scene data alliance with 12 founding members and released a data security initiative that outlines collection protocols and guarantees end-to-end protection. The platform is currently operational in 40 cities.