MotivEat: A Web-Based Meal Planning Application with Ingredient Recognition for Personalized Nutritional Guidance

Noncommunicable diseases are prevalent in the Philippines. Filipinos increasingly suffer from undernutrition or overnutrition. This can be caused by a lack of budget and nutritional awareness. Tolerating improper macronutrient intake is related to a higher risk of chronic diseases. Although varying solutions have been attempted, there is potential for discoveries. Specifically, with food choice technologies, motivation is an under-researched topic. Thus, this study explores the topic through the development of a meal planner web application that enables users to plan meals based on recipe selections tailored to their recommended macronutrient intake and available ingredients. Considering price, convenience, and experience, a proposed motivational concept based on nudging was to enable users to find recipes by scanning available ingredients. For this, a machine learning object detection model was trained to detect ingredients and find recipes. The model had a mean Average Precision (mAP) of 0.526 and an Average Recall (AR) of 0.539. Still, it could detect ingredients even when set at an accuracy of 90% and above. It was then integrated into a web application. Once accessible, the deployed web application was evaluated by nutritionist dietitians and other respondents. An average System Usability Scale (SUS) score of 79/100 was obtained. Moreover, the computed scores for the Intrinsic Motivation Inventory (IMI) subscales were all higher than 5 out of 7, with a Cronbach’s Alpha value of 0.7. Responses to open-ended questions also showed that users are motivated to eat healthier if they have increased nutritional awareness and pre-existing ingredients in suggested recipes.

Keywords: Machine learning, object detection, TensorFlow, meal planner, web application