AI Dataset2026
Provides 90,000 hours of head-mounted egocentric video paired with synchronized 3D hand pose and an optional 3D full‑body pose add-on, with event-level semantic labels available as a complimentary layer — designed for embodied AI and robotics training at scale.