AI-based calorie and macronutrient estimation for Lebanese dishes using single-view images

Abstract

This dataset is used for the study "AI-based calorie and macronutrient estimation for Lebanese dishes using single-view images". It contains image data used for food classification and for the evaluation of the proposed system under real-world conditions.

The dataset includes: (1) a food-classification dataset containing web-collected food images and synthetic food images generated using a GAN, Gemini, and Sora. The synthetic data are organized according to the study’s 174 food classes; (2) a final real-world test set containing 225 food images obtained from wearable-camera recordings collected from children in free-living settings in Lebanon as part of the SCALE study; and (3) an Excel file containing the corresponding food-class labels and ground-truth portion weights for the 225 wearable-camera images.