- Published on
Researchers from the National University of Singapore have developed a generative video pipeline that converts third-person human activity footage into synthetic training data for humanoid robots, effectively bridging the embodiment gap at scale.