Improving Image Captioning with Better Use of Captions

TL;DR

This paper introduces a novel image captioning architecture that leverages caption-guided visual relationship graphs and multi-task learning to improve image understanding and caption quality, achieving state-of-the-art results on MSCOCO.

Contribution

The paper presents a new architecture that better exploits caption semantics through visual relationship graphs and multi-task learning, enhancing both image representation and caption generation.

Findings

Significant performance improvement over baselines.

Achieved state-of-the-art results on MSCOCO.

Effective use of caption-guided visual relationship graphs.

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.