Enhancing Plant Leaf Disease Detection: Integrating Mobilenet With Local Binary Pattern And Visualizing Insights With Grad-Cam

Plant leaf disease detection plays a crucial role in agricultural management; this work presents a methodology for precise identification of plant leaf diseases, crucial for effective agricultural management and timely intervention. The proposed approach combines the efficiency of MobileNet, a lightweight convolutional neural network (CNN), with the discriminative power of Local Binary Pattern (LBP) to enhance feature extraction for accurate disease detection. The integration of LBP augments the network's capability to capture essential textural information. Additionally, for improved interpretability, Grad-CAM (Gradient-weighted Class Activation Mapping) is utilized for visualization, highlighting significant regions in input images that influence the model's predictions. This not only aids researchers in validating decisions but also provides transparent insights for farmers and practitioners in plant health monitoring. Evaluation on a comprehensive dataset demonstrates the effectiveness of the approach, achieving impressive results with 96% accuracy, 90% precision, 89% recall, and 89% F1 score. The combination of MobileNet with LBP, along with Grad-CAM, emerges as a robust tool for real-world applications in precision agriculture, fostering trust and informed decision-making in plant disease management.