Enhanced Rice Leaf Disease Detection Using Densenet201: A Deep Learning Approach

Abstract

Accurate and prompt detection of rice leaf diseases is vital for ensuring healthy crop yields and preventing major agricultural losses. This study proposes an advanced method for diagnosing rice leaf diseases using the DenseNet architecture, a leading deep learning model acclaimed for its efficiency and superior performance. Our approach employs the DenseNet-201 model to classify Ten rice leaf diseases using an extensive dataset of 15,050 images. The proposed model exhibits exceptional accuracy and robustness, achieving a validation accuracy of 98.5% and a macro average F1-score, precision, and recall of nearly 0.99. The DenseNet-201 model’s architecture, characterized by densely connected convolutional networks, facilitates efficient feature reuse and alleviates the vanishing gradient problem, enhancing its performance. We trained the model over multiple epochs, monitoring accuracy and loss metrics to ensure optimal learning and generalization. Our results indicate that the DenseNet-based approach greatly surpasses traditional machine learning methods, providing a dependable tool for farmers and agronomists to identify and diagnose rice leaf diseases early. The model's high precision and recall scores across all disease categories highlight its capability to accurately distinguish between different disease types with minimal misclassification. To revolutionize agricultural practices by introducing automated, high-precision disease detection systems. Future work will focus on deploying this model on mobile and edge devices to enable real-time disease monitoring and management in the field, ultimately contributing to sustainable agricultural productivity and food security.