Abstract: The lack of data and the difficulty of multimodal fusion have always been challenges for multimodal emotion recognition (MER). In this paper, we propose to use pre-trained models as upstream ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results