Image Processing for Classification of Rice Varieties with Deep Convolutional Neural Networks

Abstract

This research applied the Deep Convolutional Neural Networks and used the VGG16 model to screen rice varieties by images. The rice varieties selected in the experiment include five varieties: KorKhor 23, Suphanburi 1, Pathum Thani 1, Chainat 1, and Hom Mali Rice 105, totaling 1,500 images. The results of the experiments and model testing showed that the accuracy obtained by training the images of rice seeds is 85%, which is highly reliable. Therefore, the model was used to develop a website that can be accessed via web browsers and mobile apps where farmers or related persons can upload rice seed images to the system so that the system can predict what variety of rice it is and according to the testing of the system, it was found that it can make an accurate forecast of rice varieties.