Machine Learning (ML) Data Residential Egocentric Video: Household Tasks | Hand-Keypoint & Camera-Head Trajectory Data | Residential Robotics

Specifications: • Volume: 50,000 hours of machine learning (ML) data • Weekly output: 1,000+ hours of new footage added per week • Environments: 16 distinct settings — 1 residential (home & daily living) and 15 non-residential (construction, food service, retail, auto repair, warehousing, personal care, and more) • Camera: Dual 1920×1080 @ 30fps global-shutter stereo cameras, 157° horizontal / 84° vertical FOV, hardware-synchronized exposure • Motion sensing: 400Hz 9-DoF IMU (accelerometer, gyroscope, magnetometer), hardware-synced to the camera clock • Audio: Stereo, synced to video and IMU • Format: mcap and LeRobot v3, H.265/HEVC encoding • Capture hardware: Proprietary wearable rig, purpose-built for embodied-AI capture • Application scenarios: Physical AI, robot foundation models, imitation learning, egocentric vision, human-activity understanding, machine learning (ML) data • Availability: Inventory & made-to-order; volume and environment mix scoped per project • Licensing: All rights and licensing cleared for training purposes, full indemnity included.