Spatial Everyday Activities (SEA) is an egocentric dataset designed for training robotic foundation models. It comprises approximately 10,000 hours of egocentric data collected by computer vision experts across a diverse range of locations in the US and EU.
