An Assessment of the Utilization of Neural Networks for the Detection and Prediction of Rice Leaf Diseases

Numerous bacterial, viral, or fungi affect rice leaves, and these illnesses greatly lower rice yield. Understanding rice illnesses of the leaves is critical to meeting the world's large population's need for rice. However, only certain image backdrops and image capture settings can be used to identify rice leaf disease. In the realm of rice leaf disease detection, convolutional neural networks (CNNs)-based models are a popular area of research. However, the current CNN-based models are only capable of learning large-scale network parameters and have a sharp decline in recognition rates on independent datasets. In this paper, we suggest a unique CNN-based model that lowers the network parameters to identify illnesses of rice leaves. Several CNN-based models have been taught to recognize five prevalent rice leaf illnesses employing a fresh dataset of 4199 photos of rice leaf disorders. With accuracy for validation of 97.35% and an initial training reliability of 99.78%, the suggested model performs best. The suggested model's efficacy is assessed using a collection of independent photos of rice leaf disease, achieving an accuracy level of 97.82% and an area around the curve (AUC) of 0.99%. These findings show how our method is more effective and efficient than the most recent CNN-based rice leaf disease identification models.