Rice Leaf Disease Detection using MobileNet Transfer Learning Model
Rahul Singh, Neha Sharma, Rupesh Gupta
Abstract
Rice is a crucial cereal crop worldwide, consumed by more than half of the global population as a primary energy source. The productivity and value of rice grains are affected by various biotic and abiotic issues, including pests, soil fertility, precipitation, temperature, and diseases caused by bacteria and viruses. Farmers invest significant time and resources in disease management, often relying on manual visual detection methods that can lead to ineffective farming practices. However, technological advancements in agriculture have led to the development of automatic identification methods for infectious organisms in rice plant leaves. The study utilized a pre-trained deep convolutional neural network known as MobilNet, implementing a transfer learning approach to identify three significant diseases that commonly affect rice plants: Bacterial leaf blight, Brown spot, and Leaf smut. To accomplish this objective, the transfer learning model underwent fine-tuning using diverse hyperparameters and attained an accuracy rate of 87% after running for 100 epochs. The goal of this research is to precisely categorize plant leaves based on their respective disease classifications, enabling early detection and preventive measures against plant diseases.