Summary
International Workshop on Smart Info-Media Systems in Asia
2021
Session Number:RS2
Session:
Number:RS2-1
Accelaration of Food Region Extraction Based on Saliency Detection
Takuya Futagami, Noboru Hayasaka,
pp.52-55
Publication Date:2021/9/20
Online ISSN:2188-5079
DOI:10.34385/proc.66.RS2-1
PDF download
Summary:
In recent year, image-based food intake assessment attracts international interests to avoid epidemic proportions of chronic diseases such as diabetes and obesity. This paper aims to improve food region extraction, which is required to estimate energy and nutritional values of the food image, based on saliency detection in terms of computational time so as to contribute to the development of image-based food intake assessment. The algorithm of the proposed method is designed to decrease resolution of the input food image and to apply fast GrabCut, which can refine the food and background regions on the basis of graph theory. Our comparison experiment, which employed 241 actual food images, demonstrated that the proposed method decreased the computational time for each image by 1.20 s, which corresponds to 55.04 % reduction rate, compared with the conventional method, on which the proposed method is based. In addition, the proposed method increased the food extraction accuracy by 0.20 %. Although there was no significant difference in the accuracy, the proposed method was effective in that it could decrease the computational time without decreasing the accuracy. Further discussion and implications are provided herein.