Paying Attention to Descriptions Generated by Image Captioning Models

TL;DR

This paper explores how human and machine-generated image descriptions align with visual saliency, proposing a saliency-boosted captioning model that improves generalization but not performance on familiar datasets.

Contribution

It introduces a saliency-boosted image captioning model and analyzes the relationship between saliency and description quality, highlighting its benefits for unseen data.

Findings

Humans mention salient objects earlier in descriptions.

Better captioning models align more with human attention.

Saliency-boosted model generalizes better to unseen data.

Abstract

To bridge the gap between humans and machines in image understanding and describing, we need further insight into how people describe a perceived scene. In this paper, we study the agreement between bottom-up saliency-based visual attention and object referrals in scene description constructs. We investigate the properties of human-written descriptions and machine-generated ones. We then propose a saliency-boosted image captioning model in order to investigate benefits from low-level cues in language models. We learn that (1) humans mention more salient objects earlier than less salient ones in their descriptions, (2) the better a captioning model performs, the better attention agreement it has with human descriptions, (3) the proposed saliency-boosted model, compared to its baseline form, does not improve significantly on the MS COCO database, indicating explicit bottom-up boosting does not help when the task is well learnt and tuned on a data, (4) a better generalization is, however, observed for the saliency-boosted model on unseen data.