量子位 (微信公众号)T2·Models
74 pts
Shanghai AI Lab's Intern Robotics team released EgoExplorer, a wearable pipeline combining five trackers, data gloves and a head-mounted camera to synchronously capture first-person video, metric-scale whole-body motion, hand joint poses and palm tactile signals without modifying the environment. It targets the data gap in loco-manipulation, where body movement and grasp must be recorded on a shared time base, filling finger-joint and tactile dimensions missing from datasets such as Ego4D and Nymeria while enabling parallel capture across sites.