Nutrient Estimation from Images of Food for Diet Management in Diabetic Patients

Due to its close relationship to health and chronic diseases, diet has a significant impact on how individuals live their daily lives. Deep-learning based food computing is developing in the meantime to offer a variety of services, such as food identification, nutrient estimation, and so forth. This research seeks to recognize food utilizing several image processing approaches, feed that information into an deep learning architecture for classification, and obtain its nutrition information. The work is entirely coded in python language on google colaboratory. The neural network used is MobileNetV2, which is a lightweight model. The model has a test accuracy of 99% after 5 fold cross validation. Based on the obtained results of this research an Android app is developed which can easily used by diabetic patients or fitness freak, health enthusiasts, and common people in society. The app predicts the food item and displays its calorie and macronutrient contents.