A semi-automated food voting classification system: Combining user interaction and Support Vector Machines

Abstract

Obesity is prevalent worldwide including UK and Ireland, affecting all demographics. Obesity can have a detrimental affect on an individual's health, which can lead to chronic conditions. Different digital interventions have enabled users to photograph food items to be identified using different feature extraction methods. In this research, we proposed a system that allows users to draw a polygon around a food item for segmentation. After segmented, the region is then classified using an automated voting system. Different features will then be extracted from the specified area. Support Vector Machines will be issued for each feature type. This system is a proof-of-concept and is designed to research the effectiveness of employing multiple feature detection algorithms to classify food images. To classify food regions a Bag-of-features (BoFs) approach will be used for each. Speeded Up Robust Features point detection and descriptors was used along with colour spatial features, and also MSER region detection with SURF. Each of these methods will have their own BoF to train an SVM. The aim of this research was to create a voting classification system that utilises each feature detection algorithm to ultimately identify the segmented food region through plurality (or majority) vote. Testing showed that the system achieved 75% accuracy when combining each feature SVM to create a voting system. The system outperforms two of the feature classifiers (SURF and MSER with SURF). LAB colour classifier slightly outperformed the voting mechanism within the developed system. In regards to future work, further development and testing would be completed through increasing the variety of food items used in the training phase and a larger test dataset would also be used.