Image captioning: a survey of methods, datasets, evaluation metrics

Abstract

This article studies an overview of techniques to generate captions for images such as image captioning based on information retrieval, based on templates and especially based on deep learning, which has brought a revolution in generating captions for photos. In addition to updating the new studies, the study also introduces datasets for training and testing the image captioning system, common metrics to evaluate the efficiency of images captioning. The conclusion of the article proposes some research directions in the field of image captioning that researchers can study further.