GQA деректер жиынтығы

AbstractVision-and-language reasoning requires an understanding of visual concepts, language semantics, and, most importantly, the alignment and relationships between these two modalities. We thus propose the LXMERT (Learning Cross-Modality Encoder Representations …...

AbstractThe predominant approach of visual question answering (VQA) relies on encoding the imageand question with a ”black box” neural encoder and decoding a single token into answers suchas ”yes” or …...