Improving Image Captioning with Better Use of Caption

Zhan Shi, Zhou Xu, Xipeng Qiu, Xiaodan Zhu

2020 · 85 citations · 32 references

DOIFull text

Open access

Concepts

TL;DR

Image captioning is a multimodal problem that has attracted extensive attention in both natural language processing and computer vision. The paper proposes a novel image captioning architecture that better exploits caption semantics to improve image representation and caption generation. The architecture builds caption‑guided visual relationship graphs via weakly supervised multi‑instance learning, enriches representations with neighboring and contextual nodes, and uses multi‑task learning during generation to jointly predict words and object/predicate tags. Experiments on MSCOCO demonstrate that the framework significantly outperforms baselines, achieving state‑of‑the‑art results across multiple metrics, and the code is publicly available.

Abstract

Image captioning is a multimodal problem that has drawn extensive attention in both the natural language processing and computer vision community. In this paper, we present a novel image captioning architecture to better explore semantics available in captions and leverage that to enhance both image representation and caption generation. Our models first construct caption-guided visual relationship graphs that introduce beneficial inductive bias using weakly supervised multi-instance learning. The representation is then enhanced with neighbouring and contextual nodes with their textual and visual features. During generation, the model further incorporates visual relationships using multi-task learning for jointly predicting word and object/predicate tag sequences. We perform extensive experiments on the MSCOCO dataset, showing that the proposed framework significantly outperforms the baselines, resulting in the state-of-the-art performance under a wide range of evaluation metrics. The code of our paper has been made publicly available.

References

32