Layer-wise Fusion with Modality Independence Modeling for Multi-modal Emotion Recognition

38816

Knowledge Triples