Enhancing Food Analysis with Attention-based Deep Learning: Ingredients, Recipes, and Calorie Estimation

The past several decades have seen a significant increase in food’s diversity and popularity as there is an increasing health consciousness regarding diet among individuals. For those who are interested in fitness, tracking calories is now crucial. Conversely, deep learning has become more popular for resolving a variety of issues related to natural language processing and computer vision. Therefore, in order to extract components, recipes, and the total number of calories from the food photographs, this research developed a novel deep learning based method. This work makes two main contributions: first, it presents a novel attention-based deep learning model that uses transfer learning to produce recipes and food components from food photos. Secondly, it offers a way for figuring up the overall caloric content by utilizing the components and amounts of the dish. The suggested model not only helps people by pointing out items that can cause allergies, but it also indicates whether a dish is vegetarian or not. According to experimental data, our model outperforms existing models by 5% improvement in the accuracy.