To understand the nutrients we consume, the first step is to record what we eat. However, keeping track of every meal and snack should be challenging. We are developing a food logging app based on deep image recognition technology called FoodLog Athl. When you upload an image of your meal, the app detects the area of each food and recognizes the dish name, allowing it to estimate the nutritional value of that meal. The users who have access to a dietitian’s support can share their food records with the dietitians and communicate via chat.

For example, even the same dish, "nikujaga," can have significantly different nutritional values depending on the recipe. We are developing a meal nutrition calculation app, RecipeLog, that estimates the types and quantities of ingredients from meal images, allowing for detailed nutritional calculations. The app features functionalities enabling users to assess their meals by visualizing how the nutritional values calculated from their created ingredient list fulfill their daily recommended intake.

Food frequency questionnaires (FFQs) are commonly used in cohort studies involving large numbers of participants. In this study, we aim to develop a system that automatically converts dietary records̶which are initially completed by participants and then supplemented by dietitians through interviews̶into a data format suitable for nutritional calculations based on standard food composition tables, by fine-tuning a large language model.