EntréeNet: A Triple Input CNN Food Classification System using Multi-color Spaces

Monitoring food intake is essential because of the link between dietary choices and their impact on health. This research addresses the need for accurate food monitoring to prevent health issues like obesity, heart disease, and diabetes, with a particular emphasis on categorizing different types of food. While current food classification models can effectively categorize food images, their approach is solely based on RGB images. Hence, it is necessary to study the impact of utilizing different color spaces in classifying food images. Using different color spaces can allow models to leverage the discriminative features of various color space representations, learn more robust features, and differentiate food images more effectively. This research presents EntréeNet, a Triple Input Convolutional Neural Network that classifies food images using several color spaces. This study proposes an alternative approach that improves on the overall accuracy of food image classification models. The results of the different experiments reveal that models utilizing several color spaces for food image classification generally perform better than models relying solely on a single color space.