FoodVisor: A Food Calorie Estimation System

In recent global studies, it has been observed that maintaining a healthy diet is crucial for overall well-being. Today, individuals are more conscious about what they eat and about their dietary choices than ever before. With the advancements in scientific research and technology, there is now the possibility of creating a personalized food identification system to monitor daily calorie intake. However, the development and implementation of such a system poses various challenges. In our research paper, we have introduced a neural network-based model designed to predict food items from images or pictures and provide estimated calorie information. To achieve this, we gathered a dataset consisting of approximately 101,000 images spanning 101 different food categories and we achieved the results namely a Recall value of 0.779, F1 score of 0.776, and Precision value of 0.777. Our approach involved training a Convolutional Neural Network (CNN) with features extracted using the ResNet50 model, which is a transfer learning model, resulting in a model with an accuracy of 77.97%. The deployment of our system is accessible through a webpage. People can upload an image of food items, and in real-time, our system will predict the identified food item along with its estimated calorie content. This project demonstrates the potential of integrating technology into dietary monitoring for a more informed and health-conscious lifestyle.