Portion Estimation

Portion estimation aims to accurately determine the quantity of food consumed, primarily using image-based methods to overcome limitations of traditional self-reporting. Current research focuses on improving accuracy by incorporating 3D information from single or multiple images, leveraging deep learning architectures like transformers and generative models to reconstruct food shapes and estimate volume, and employing novel data acquisition techniques such as tracking food on utensils. These advancements hold significant promise for improving dietary assessment, enabling more precise calorie counting, and facilitating personalized nutrition recommendations.