A Deep Learning Approach For Automated Rice Disease Detection And Classification

Rice is considered one the most important plants globally because it is a source of food for over half the world's population. Like other plants, rice is susceptible to diseases that may affect the quantity and quality of produce. Because of this, crops sometimes lose 20% to 40% of their value. A good harvest can depend on early discovery of these diseases, so farmers would need to be able to read and understand images of the diseases. Also, farmers still can't reach their goal of doing daily studies of their huge farmlands. It would be very expensive to do this, so the price of rice for buyers would go up even if it were possible. This paper proposed a pre-trained Deep Convolutional Neural Network (DCNN) method based on optimization for accurately finding and classifying rice leaf disease. It uses both transfer learning and baseline learning. A precise diagnosis method can find and classify eleven different types of rice diseased healthy, leaf blast, brown spot, bacterial blight and bacterial leaf blight, false stump, neck blast, stemborer, tumgro, hispa, and BPH. The most advanced large-scale architecture, such as XceptionNet, ResNet50, DenseNet VGG19, SequeezeNet and CNN used for recognition of the Rice disease, with SGDM, ADAM, RMS propagation optimization methods for predictions for a dataset. The -proposed models were trained and tested using datasets gathered from websites. In the simulation results consistently demonstrate that the XceptionNet model outperforms other architectures in terms of higher accuracy 93.3 %.