AGIC: Attention-Guided Image Captioning to Improve Caption Relevance

TL;DR

AGIC introduces an attention-guided approach with hybrid decoding to enhance caption relevance, achieving state-of-the-art results with faster inference in image captioning tasks.

Contribution

This paper presents a novel attention-guided method and hybrid decoding strategy that improve caption relevance and inference speed in image captioning.

Findings

AGIC outperforms several state-of-the-art models.

AGIC achieves faster inference times.

Strong performance across multiple metrics.

Abstract

Despite significant progress in image captioning, generating accurate and descriptive captions remains a long-standing challenge. In this study, we propose Attention-Guided Image Captioning (AGIC), which amplifies salient visual regions directly in the feature space to guide caption generation. We further introduce a hybrid decoding strategy that combines deterministic and probabilistic sampling to balance fluency and diversity. To evaluate AGIC, we conduct extensive experiments on the Flickr8k and Flickr30k datasets. The results show that AGIC matches or surpasses several state-of-the-art models while achieving faster inference. Moreover, AGIC demonstrates strong performance across multiple evaluation metrics, offering a scalable and interpretable solution for image captioning.