Multimodal Transformer With a Low-Computational-Cost Guarantee

Sungjin Park, Edward Choi

Transformer-based models have significantly improved performance across a range of multimodal understanding tasks, such as visual question answering and action recognition. However…