Empirical Method for Rice Grain Variety Classification

Rice is grown extensively worldwide and displays a wide genetic diversity., resulting in a variety of traits and characteristics across different types. In order to achieve realistic Rice Grain Classification., this research compared the use of Vgg19., MobileNet., MobileNet v2., CNN and DensNet algorithms. The data set used in this study collected from various data repositories such as kaggle.,GitHub.,VNMN and etc. The dataset comprises a total of 10.,000 high-resolution images of rice grains., with each image labeled according to its corresponding rice variety. The dataset included five distinct rice varieties: Akura., 1001., Konaram., MTU-64., and RNR. Each rice variety is represented by 2.,000 images in the dataset. Then CNN., Vgg19., MobileN et., MobileN et v2 and DensN et are trained and assessed using the enhanced dataset. Our research results revealed successful result with DenseN et emerging as the top performer., achieving an impressive accuracy of 96.87%. Following closely behind., MobileNet V2 demonstrated notable proficiency with a classification accuracy of 94.60%., while MobileNet and VGG 19 attained accuracies of 94 % and 90%., respectively.