Object Relation Attention for Image Paragraph Captioning

Image paragraph captioning aims to automatically generate a paragraph from a given image. It is an extension of image captioning in terms of generating multiple sentences instead of a single one, and it is more challenging because paragraphs are longer, more informative, and more linguistically complicated. Because a paragraph consists of several sentences, an effective image paragraph captioning method should generate consistent sentences rather than contradictory ones. It is still an open question how to achieve this goal, and for it we propose a method to incorporate objects' spatial coherence into a language-generating model. For every two overlapping objects, the proposed method concatenates their raw visual features to create two directional pair features and learns weights optimizing those pair features as relation-aware object features for a language-generating model. Experimental results show that the proposed network extracts effective object features for image paragraph captioning and achieves promising performance against existing methods. rich details depicted by an image. To address this problem, Krause et al (Krause et al. 2017 ) compile a dataset containing thousands of images and their corresponding paragraphs in 5 to 8 descriptive sentences. This is the only dataset designed for image paragraph captioning to the best of our knowledge, which aims to generate multiple sentences as an integrated description of an image. Since image paragraph captioning is extended from image captioning, a straight-forward extension for image paragraph captioning is to generate sentences after sentences to form a paragraph. However, image captioning approaches