Tag: multimodal transformer

A new supervised cross-modal hashing method called SEGRH converts label co-occurrence into continuous semantic association scores and reports state-of-the-art retrieval ...

A new hybrid multimodal transformer framework combines contrastive annotation, cross-modal attention, graph-based lexical expansion, and ordinal hate-intensity scoring, reporting 91.8 ...

A new multimodal Transformer framework called MFT-Net dynamically re-weights clicks, speech, gestures, and facial expressions to recognize 18 categories of ...