Food Image Recognition and Volume Estimation: A Comprehensive Study for Dietary Assessment

The paper discusses the importance of health and wellness in today’s tech-dependent society and introduces the model named as "Food Diet Recaller App - FDRA". This app employs AI and computer vision to help users track and manage their diets effectively. It features food recognition, quantity estimation, and provides detailed nutritional information. The research utilizes techniques like image recognition and volume estimation, to accurately measure food intake. It also incorporates the InceptionV3 model for image classification an impressive accuracy of approximately 92% is achieved. The Ninja API offers nutritional information for classified food items. With the reference of aruco marker users capture images of their food, and the app calculates volume and density to determine nutritional content. The food volume estimation model is specifically trained for linear-shaped food items, characterized by square or rectangular surfaces. Additionally, the development of a dataset and regression models is mentioned to improve accuracy in estimating food mass. The paper also highlights the popularity of diet recall apps and their role in helping users maintain balanced nutrition.