A Comprehensive Survey on Image Captioning Techniques

Image captioning technology is an intersection of computer vision and natural language processing, which utilizes computer vision-related techniques to extract semantic information embedded in images, and then transforms this semantic information into human-understandable text descriptions through natural language processing-related techniques, which promotes innovations in the fields of humancomputer interaction, accessibility assistive technology, and intelligent multimedia retrieval. In this paper, according to the time course of the development of image captioning technology, the technology is classified into template-based methods, retrievalbased methods, and deep learning-based methods according to the different image description methods proposed in 21 literatures, and these methods are introduced and summarized in turn, followed by the introduction of commonly used scoring metrics and datasets in image captioning technology.