A Deep Convolutional Neural Network for Food Detection and Recognition

Abstract

In this paper, we propose a new deep convolutional neural network (CNN) configuration to detect and recognize local food images. Various types of food with different color and texture reflect the fact that the food image recognition is considered a challenging task. However, deep learning has been widely used as an efficient image recognition method, and CNN is the contemporary approach for deep learning to be implemented. CNN has been optimized to the tasks of food detection and recognition with few modifications. We present a new dataset of the most consumed local Malaysian food items which was collected from publicly available Internet sources including but not limited to, image search engines. For evaluation of recognition performance, CNN achieved significantly higher accuracy than traditional approaches with manually extracted features. Additionally, it was found out that convolution masks show that the features of food color dominate the features map. For the process of food detection, CNN also exhibited considerably higher accuracy than other conventional methods.