Fast and Accurate Calorie Count Prediction from Food Images using a Convolution Neural Network Method

Abstract

In the past ten years, with advances in deep learning techniques, automated object recognition has come very near to human levels of accuracy. Fast, automatic, and consistent image-based food calorie calculation is becoming a requirement with the increasing growth of overweight and other lifestyle- related disorders throughout the globe. Accurate and insightful solutions may be provided in the form of a mobile app with the aid of a deep learning-based automatic object recognition system. However, real-time image processing necessitates a high level of processing speed, which is a major consideration for such applications. Although several researchers have looked at estimating calories based on pictures of food, there is currently no image-driven, lightweight, rapid, and accurate food calorie estimation method. In this study, we offer a technique for recognizing common meals captured with a mobile phone camera by using Convolution Neural Networks (CNN) with optimum parameters. Calories and other nutritional information may be deduced from the known food class after the food items have been identified. Our research shows that our suggested method not only guarantees precision but also has the potential to greatly streamline the complex, time-consuming, and labor-intensive processes now used for estimating calorie intake by turning them together into real-time automation techniques.