Automatic Prediction of Glycemic Index Category from Food Images Using Machine Learning Approaches

The dietary glycemic index (GI) is a mechanism, whereby carbohydrate-containing foods are assigned a number according to how much each food raises blood sugar. It plays a very important role in the health of a person as this affects significantly metabolic disorders especially when counts for diabetes and obesity. Serious interventions are therefore required for GI in terms of medical perspective. There are a series of breakthroughs, which attempt to improvise the classification of food types from food images using machine learning (ML) techniques. However, no research is reported for the classification of food type based on the food GI. Therefore, the proposed work investigates food image classification based on their GI index category. The work is the first of its kind. The proposed framework is employed on food images from foodpics_extended databases that are based on international GI tables of three food categories for low, medium, and high. An extensive range of texture, statistical, and shape features was extracted from the images. Thereafter, experiments were performed with five different categories of classifiers, viz. AdaBoost with random forest (RF), J48 decision tree, k-nearest-neighbor (KNN) classifier, Naive Bayes classifier, and sequential minimal optimization (SMO)-based support vector machine (SVM) classifier. Statistical analysis was conducted to compare the effectiveness of the different methods. Though no method performs significantly better than others, we observed that the overall performance of the AdaBoost (RF) ensemble model provided better classification accuracy in finding the category of food image GI index.