Object Relation Attention for Image Paragraph Captioning

Image paragraph captioning aims to automatically generate a from given image. It is an extension of image in terms generating multiple sentences instead single one, and it more challenging because paragraphs are longer, informative, linguistically complicated. Because consists several sentences, effective method should consistent rather than contradictory ones. still open question how achieve this goal, for we propose incorporate objects' spatial coherence into language-generating model. For every two overlapping objects, the proposed concatenates their raw visual features create directional pair learns weights optimizing those as relation-aware object Experimental results show that network extracts achieves promising performance against existing methods.