CLASSIFICATION OF RICE TEXTURE BASED ON RICE IMAGE USED THE CONVOLUTIONAL NEURAL NETWORK METHOD

Abstract

There are several types of rice that are commonly sold in rice stores. Many people, especially millennials, are not familiar with the different types of rice such as IR42 rice, Pera rice, sticky rice, and Pandan Wangi rice. Therefore, digital image processing techniques are needed to help analyze the types of rice to help people know what kind of rice they are going to buy at the market. The method commonly used in image processing for image classification is the convolutional neural network (CNN) method. Currently, CNN has shown the most significant results in image classification. This research used a dataset of 1560 rice images. The data was divided into two sets (training data and validation data) with an 80:20 ratio. The accuracy obtained by the CNN model using InceptionV3 for the rice data was 95.7% with a loss of 0.123. The Android application developed in this research achieved an accuracy of 83,4% based on the testing results calculated using the confusion matrix.