Smart nutrition monitoring system using serverless edge computing

Diet-related health problems, such as obesity, diabetes and heart disease have grown to epidemic proportions in the world and are taking a heavy toll on our society and our healthcare systems. One crucial factor contributing to these problems is the lack of proper measurement and control of food intake. In this paper, we address this gap by proposing a new smart food scanner based on emerging serverless edge computing. The proposed smart scanner is a multi-sensor appliance equipped with a camera, scale and edge devices with the ability to connect to smartphones to enable users to measure and analyse their food intake and support nutritional decision-making. We developed a food recognition workflow using different deep learning models for food segmentation, classification and food volume estimation with a high level of accuracy. The proposed smart nutrition monitoring system has been implemented and evaluated using a real food dataset which reveals that we can estimate the nutritional values with an average error of 6.1%. Using serverless edge computing architecture, the system latency is about 10 seconds for scanning the food and about 110.26 seconds for reporting the results to users via the mobile app.