Robot Imitation Learning · World Model Training · Cross-Embodiment Unified Data
In support of LeRobot 3.0 format and JSON metadata delivery, high-quality data are provided for the 2026 front-line demand for smart and world models, including Ego-centric human behaviour, industrial assembly, commercial retail, logistics sorting, home service, office documentation, catering services, etc.
Cross-scenario model human operation data, supporting behavioral cloning, ACT, Diffusion Policy, and other learning algorithms.
Long video and multimodular sensor data build a world model that is physically and predictably interactive.
The uniform metadata format is agreed upon with the coordinates system, supporting training in mechanical arm, human-shaped robots, mobile robots and sharing.
Support for the mainstream training framework formats for LeRobot 3.0, RLDS, HDF5, JSON metadata, etc.
When the deliverable is egocentric video, hand motion, or teleoperation, device choice sets sync, scale, and reproducibility. Open the on-site matrix. Specification pages are published in Chinese; the matrix follows this language when a translation exists.
Capture hardware Ego · gloves · teleop Open the device matrix →Changsha Langhui Information Technology Co., Ltd. The DataAssetsAPI platform provides AI with high-quality, compliant and smart data assets for enterprises and scientific institutions.
Contact Now