Сетки Активации

AbstractThis paper studies the problem of generating a summary for a given sentence-image pair. Existing multimodal sequence-to-sequence approaches mainly focus on enhancing the decoder by visual signals, while ignoring that …...