[논문리뷰 : 개념] VATT : Transformers for Multimodal Self-Supervised Learning From Raw Video, Audio and Text