Goldilocks and the Three Parameters:Empirically Finding the "Just Right" for Segmenting Food Images for the AFINI-T System

Abstract

Measuring nutritional intake is a tool that is critical to the
monitoring of health, both as an individual or of a group. It is
especially important in the monitoring of those at risk for
malnutrition, an issue which costs billions of dollars globally, and
current methods used in practice are manual, time-consuming,
and have inherent biases and inaccuracies. This study proposes a
novel imaging system with a superpixel-based segmentation
algorithm as part of an automated nutritional intake system. The
study also examines three important parameters of the algorithm
and their ideal values; region size and spatial regularization for
superpixel segmentation, as well as spatial weighting in
clustering. The experimental results demonstrate that the
proposed system is effective in segmenting an image of a plate into
its constituent foods.