Component Based Comparative Analysis Of Each Module In Image Captioning

Image captioning is a task to generate a new caption using the training data of the image and caption. Since existing deep learning is a black-box model, it is crucial to analyze the influence on each module for understanding the model. In this paper, we analyze the impact of the five modules and do a comparative analysis according to three losses and two optimizations using two datasets. From extensive experiments, the best component of each module has been identified as an improved method. (C) 2021 The Korean Institute of Communications and Information Sciences (KICS). Publishing services by Elsevier B.V.