A Deep Learning Neural Network-based System for Food Recognition and Calorie Estimation

The global increase in health awareness reflects a rising desire among individuals to pursue healthy and normal lives. Obesity has emerged as an important health issue in the modern world due to its rapid pace. Obesity is defined by medical authorities as having a Body Mass Index (BMI) of more than 30 kg/m2, which can lead to a variety of health difficulties like liver failure, cholesterol, respiratory disorders, diabetes, cardiac complications, and, in serious cases, cancer. Maintaining a diet rich in nutrients and low in calories is critical in the fight against obesity. Individuals, on the other hand, frequently struggle with food decisions. In response to this difficulty, this research presents a system that offers calorific calculation followed by food recognition to aid in preventing obesity. The goal of this research is to identify types of food and estimate their calorific values. For food recognition, a novel Attention-based Dilated Convolutional Neural Network Logistic Regression (A-CNNLR) model is introduced and compared to a traditional CNN model. The proposed ACNNLR model demonstrates a recognition rate of 99%, surpassing the CNN model’s 97%. The study determines the calories in the food according to their volume using data from the ECUST Food Dataset. This study provides a novel approach to the fight against obesity by using sophisticated algorithms for precise food detection and calorie estimation.