DeepFood: Deep Learning-Based Food Image Recognition for Computer-Aided Dietary Assessment

摘要

Worldwide, in 2014, more than 1.9 billion adults, 18 years and older, wereoverweight. Of these, over 600 million were obese. Accurately documentingdietary caloric intake is crucial to manage weight loss, but also presentschallenges because most of the current methods for dietary assessment must relyon memory to recall foods eaten. The ultimate goal of our research is todevelop computer-aided technical solutions to enhance and improve the accuracyof current measurements of dietary intake. Our proposed system in this paperaims to improve the accuracy of dietary assessment by analyzing the food imagescaptured by mobile devices (e.g., smartphone). The key technique innovation inthis paper is the deep learning-based food image recognition algorithms.Substantial research has demonstrated that digital imaging accurately estimatesdietary intake in many environments and it has many advantages over othermethods. However, how to derive the food information (e.g., food type andportion size) from food image effectively and efficiently remains a challengingand open research problem. We propose a new Convolutional Neural Network(CNN)-based food image recognition algorithm to address this problem. Weapplied our proposed approach to two real-world food image data sets (UEC-256and Food-101) and achieved impressive results. To the best of our knowledge,these results outperformed all other reported work using these two data sets.Our experiments have demonstrated that the proposed approach is a promisingsolution for addressing the food image recognition problem. Our future workincludes further improving the performance of the algorithms and integratingour system into a real-world mobile and cloud computing-based system to enhancethe accuracy of current measurements of dietary intake.