Food Recognition using Ingredient-Level Features

Food recognition is a difficult problem, because unlike objects like cars, faces, or pedestrians, food is deformable and exhibits high intra-class variation. This paper considers the approach of analyzing a food item at the pixellevel by classifying each pixel as a certain ingredient, and then using statistics and spatial relationships between those pixel ingredient labels as features in an SVM classifier. We experimented with multiple variations on past methods, and found that using pixel ingredient labels to identify food greatly increases classification accuracy, but at the expense of higher computational cost.