Detecting multiple rice diseases using transfer learning CNN method

Detecting multiple rice diseases is critical for sustaining agricultural productivity and food security, particularly in rice- dependent nations like the Philippines. Traditional manual disease detection methods are time-consuming and prone to errors due to overlapping symptoms across diseases. This study leverages the ResNet50 convolutional neural network (CNN) architecture, known for its deep learning capabilities and efficient residual connections, to classify 14 rice diseases with remarkable accuracy. By incorporating transfer learning and image augmentation techniques, the model achieved a classification accuracy of 99%, outperforming other architectures like MobileNet and EfficientNet, which attained accuracies of 87% and 91%, respectively. The results highlight the efficacy of ResNet50 in handling complex datasets, particularly in distinguishing diseases with overlapping symptoms. This automated approach offers significant potential to improve disease management, reduce crop losses, and enhance agricultural sustainability in the Philippines and other rice-producing regions.