Learning to Guide Decoding for Image Captioning

TL;DR

This paper introduces a guiding network to enhance image captioning by modeling image attributes and integrating this information into the decoder, leading to improved caption quality.

Contribution

It proposes a novel guiding network that can be integrated into encoder-decoder models and trained end-to-end to improve image captioning performance.

Findings

Improved caption quality on MS COCO dataset

Effective modeling of image attributes for captioning

End-to-end trainable guiding mechanism

Abstract

Recently, much advance has been made in image captioning, and an encoder-decoder framework has achieved outstanding performance for this task. In this paper, we propose an extension of the encoder-decoder framework by adding a component called guiding network. The guiding network models the attribute properties of input images, and its output is leveraged to compose the input of the decoder at each time step. The guiding network can be plugged into the current encoder-decoder framework and trained in an end-to-end manner. Hence, the guiding vector can be adaptively learned according to the signal from the decoder, making itself to embed information from both image and language. Additionally, discriminative supervision can be employed to further improve the quality of guidance. The advantages of our proposed approach are verified by experiments carried out on the MS COCO dataset.