wi-fi-sensing

CSI-JEPA is a self-supervised framework for learning reusable representations from unlabeled Wi-Fi channel state information, enabling label-efficient multi-task sensing. It achieves up to 98% label savings and outperforms supervised models.