Rice Leaf Disease Classification Using Deep Neural Networks

Rice is a staple food for over half the world, and rice is a key food security ingredient in Nigeria. However, the foliar diseases like bacterial blight, brown spot, and leaf smut may reduce rice production, which may result in a loss of up to 50%. Early detection of such disorders is hence necessary, but farmers in the remote areas have a significant problem because they have little access to specialists. The conventional laboratory testing and the practice of visual inspection of crops are usually cost-prohibitive and time-consuming for rural farmers. The use of deep learning to conduct a computer-based analysis of the disease state of rice leaves is examined to overcome these limitations in the current study. In Gombe State, Nigeria, a dataset of 860 images was obtained, and it consists of three categories, which contain infected and healthy leaves. The images have been significantly pre-processed and used to train a deep neural network model through transfer learning using EfficientNetB0.The model achieved a training accuracy of 99.8% after 50 epochs. However, the held-out test dataset accuracy was 92.7%, indicating strong generalization performance. The macro-average F1-score of 0.93 demonstrates reliable discrimination between infected and healthy leaves across all classes. As a comparison, the proposed model produces better results than the previously reported CNN model, which produced an accuracy of 95 percent. In addition to this, the presented strategy is more efficient. Deep neural networks transfer learning can help quickly, effectively, and accurately detect diseases, which would impact the food security in Nigeria through timely interventions. This is shown in the current research.