Rice Grain Classification Using Convolutional Neural Networks

Abstract

This research presented the classification of three popular rice varieties in Chiang Rai province, namely Pathum Thani 1 rice, RD 6 rice, and Japanese rice, using a convolutional neural network. This research collects images of rice grains from mobile cameras. The images were then processed into three different image formats: color
images, grayscale images, and black and white images. Then the obtained rice grain images were processed with a convolutional neural network using TensorFlow - Keras Sequential model. This research compared the efficiency of rice seed classification using three different image preparation methods: color images, grayscale images, and black and white images, and used a total of 600 rice grain images. The experiment found that learning from grayscale images achieved the highest accuracy, which was 87.22%, followed by color images and black and white images, which were 85.55% and 81.66%, respectively.