Development of android-based interface to determine color additives in food embedded with convolution neural networks technique

Abstract

Recent advanced technology enables Android smartphone suitable for quality evaluation of food. In this research, image processing technique was used to detect food color additives. In this research, a smartphone application was developed to determine the availability of color additives in food products. Local food namely geplak was made by adding food grade (i.e. tartrazine and erythrosine) and non-food grade (Rhodamin B and Methanyl Yellow) additives in three concentrations. A mobile phone captured geplak images resulting 1200 images which were divided into 1000 images for training and 200 images for validation. Image data was processed with the python programming language of tensorflow function. The output of python in nominal weight was then trained and tested by using a convolutional neural networks (CNN) method. The weights were then processed by Android Studio version 3.2.1 using.java as backend from CNN and.xml as an application layout. Validation result showed that the program successfully determined class of food additive in high degree accuracy of 98 %.