An Advanced Deep Learning Framework for Proactive Potato Leaf Disease Detection to Revolutionize Agriculture

Abstract

Potatoes, critical to global food security, are highly susceptible to various diseases, particularly those that affect leaf health. Traditional methods for diagnosing potato leaf diseases frequently rely on manual inspection, which is time-consuming and susceptible to human error. Recent advances in machine learning and profound learning algorithms, such as Convolutional neural networks (CNNs), have shown promise for automating disease detection. This study proposes a more advanced deep-learning framework for proactively detecting potato leaf diseases. The CNN model is trained on various potato leaf images captured using in-field imaging techniques to simulate real-world conditions. This method addresses dataset diversity and symptom complexity issues, resulting in significantly higher detection accuracy. The proposed framework demonstrates superior performance compared to conventional methods, achieving a remarkable diagnostic accuracy of $99.3 \%$ with AlexNet, 94.4% with VGG19, 90.1% with CNN, 89.9% with ResNet50, and $94.0 \%$ with InceptionV3. These accuracies were obtained over 20 epochs, with validation accuracies of $98.4 \%$, $96.2 \%, 94.7 \%, 47.6 \%$, and $93.0 \%$ respectively. Our framework seamlessly integrates agricultural technology and digitalization, allowing for early disease identification and categorization. This proactive approach is critical for maximizing agricultural production, increasing crop yield, and reducing losses. Furthermore, we clarify that our goal of improving detection accuracy is consistent with agrarian sustainability by encouraging timely disease intervention and reducing the need for excessive pesticide use. The proposed CNN architecture achieves high diagnostic performance by integrating innovative farming practices, demonstrating the potential for real-time disease control in agricultural settings.

Topics & Concepts