Nasi Food Image Classifier: Automated Classification of Multiple Nasi Labelled Images (i.e. rice) using Deep Learning Methods

This paper is about how to use deep learning to classify nasi (i.e. rice; Malaysia term: nasi), the paper contains 6 parts, introduction about deep learning, nasi, and how to use deep learning to classify six types of nasi. Literature survey about deep learning, and how to proposed deep learning on recognition for nasi(this contains CNN architecture, VGG16, Xception and ResNet transfer learning model, the dataset which used to train the model and the augmentation, Xception is the best model, it has the highest validation accuracy 96.9% while ReNet has highest test accuracy 96.2%.), the performance result(which contains effects of hyper-parameters of the proposed model, the plot of loss and accuracy of the model, the comparision of classification accuracies of proposed and transfer model ), the conclusion of this project and the future of this model. The basic and the most important thing in this project is to collect dataset and make a classification for nasi which is a type of food that consists mainly of rice. The nasi in this project consists of six types (Nasi briyani, Nasi dagang, Nasi kerabu, Nasi lemuni, Nasi minyak, Nasi tomato, each type of nasi has about 300 pictures, and there are 1800 pictures in total). We consider these six different kinds of nasi as label in this project.